2026-02-27 07:32:24 +00:00
21 changed files with 3516 additions and 120 deletions
--- a/backend/app/api/endpoints/ai_builder.py
+++ b/backend/app/api/endpoints/ai_builder.py
@@ -10,7 +10,6 @@
 import logging
 from typing import Annotated

-import anthropic
 from fastapi import APIRouter, Depends, HTTPException, Request, status
 from sqlalchemy.ext.asyncio import AsyncSession

@@ -52,10 +51,9 @@ def _require_ai_enabled() -> None:
    if not settings.ai_enabled:
        raise HTTPException(
            status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
-            detail="AI flow builder is not configured. Set ANTHROPIC_API_KEY.",
+            detail="AI flow builder is not configured. Set GOOGLE_AI_API_KEY or ANTHROPIC_API_KEY.",
        )

-
@router.get("/quota", response_model=AIQuotaStatusResponse)
 async def get_quota(
    current_user: Annotated[User, Depends(get_current_active_user)],
@@ -174,27 +172,6 @@ async def scaffold(
        branches, input_tokens, output_tokens, cost = await scaffold_branches(
            conversation.wizard_state,
        )
-    except anthropic.APIError as e:
-        await record_ai_usage(
-            user_id=current_user.id,
-            account_id=current_user.account_id,
-            conversation_id=conversation.id,
-            generation_type="scaffold",
-            tier=plan,
-            input_tokens=0,
-            output_tokens=0,
-            estimated_cost=0,
-            succeeded=False,
-            counts_toward_quota=False,
-            error_code=type(e).__name__,
-            extra_data={"error": str(e)},
-            db=db,
-        )
-        await db.commit()
-        raise HTTPException(
-            status_code=status.HTTP_502_BAD_GATEWAY,
-            detail="AI provider error. Please try again.",
-        )
    except ValueError as e:
        await record_ai_usage(
            user_id=current_user.id,
@@ -216,6 +193,28 @@ async def scaffold(
            status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
            detail=f"AI returned invalid output: {e}",
        )
+    except Exception as e:
+        logger.exception("AI scaffold failed: %s: %s", type(e).__name__, e)
+        await record_ai_usage(
+            user_id=current_user.id,
+            account_id=current_user.account_id,
+            conversation_id=conversation.id,
+            generation_type="scaffold",
+            tier=plan,
+            input_tokens=0,
+            output_tokens=0,
+            estimated_cost=0,
+            succeeded=False,
+            counts_toward_quota=False,
+            error_code=type(e).__name__,
+            extra_data={"error": str(e)},
+            db=db,
+        )
+        await db.commit()
+        raise HTTPException(
+            status_code=status.HTTP_502_BAD_GATEWAY,
+            detail=f"AI provider error ({type(e).__name__}). Please try again.",
+        )

    # Record successful usage
    await record_ai_usage(
@@ -293,27 +292,6 @@ async def branch_detail(
                existing_branches,
            )
        )
-    except anthropic.APIError as e:
-        await record_ai_usage(
-            user_id=current_user.id,
-            account_id=current_user.account_id,
-            conversation_id=conversation.id,
-            generation_type="branch_detail",
-            tier=plan,
-            input_tokens=0,
-            output_tokens=0,
-            estimated_cost=0,
-            succeeded=False,
-            counts_toward_quota=False,
-            error_code=type(e).__name__,
-            extra_data={"error": str(e), "branch_name": data.branch_name},
-            db=db,
-        )
-        await db.commit()
-        raise HTTPException(
-            status_code=status.HTTP_502_BAD_GATEWAY,
-            detail="AI provider error. Please try again.",
-        )
    except ValueError as e:
        await record_ai_usage(
            user_id=current_user.id,
@@ -335,6 +313,28 @@ async def branch_detail(
            status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
            detail=f"AI returned invalid output: {e}",
        )
+    except Exception as e:
+        logger.exception("AI branch_detail failed: %s: %s", type(e).__name__, e)
+        await record_ai_usage(
+            user_id=current_user.id,
+            account_id=current_user.account_id,
+            conversation_id=conversation.id,
+            generation_type="branch_detail",
+            tier=plan,
+            input_tokens=0,
+            output_tokens=0,
+            estimated_cost=0,
+            succeeded=False,
+            counts_toward_quota=False,
+            error_code=type(e).__name__,
+            extra_data={"error": str(e), "branch_name": data.branch_name},
+            db=db,
+        )
+        await db.commit()
+        raise HTTPException(
+            status_code=status.HTTP_502_BAD_GATEWAY,
+            detail=f"AI provider error ({type(e).__name__}). Please try again.",
+        )

    # Record successful usage
    await record_ai_usage(
--- a/backend/app/api/endpoints/ai_fix.py
+++ b/backend/app/api/endpoints/ai_fix.py
@@ -0,0 +1,78 @@
+"""AI auto-fix endpoint for tree validation errors.
+
+POST /ai/fix-tree — accepts a tree with validation errors and returns
+AI-generated fix proposals for each error.
+"""
+
+import logging
+from typing import Annotated
+
+from fastapi import APIRouter, Depends, HTTPException, Request, status
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.api.deps import get_db, require_engineer_or_admin
+from app.core.config import settings
+from app.core.rate_limit import limiter
+from app.core.ai_fix_service import generate_fixes
+from app.models.user import User
+from app.schemas.ai_fix import (
+    AIFixTreeRequest,
+    AIFixTreeResponse,
+    AIFixProposal,
+    AIFixTokenUsage,
+)
+
+logger = logging.getLogger(__name__)
+
+router = APIRouter(prefix="/ai", tags=["ai-fix"])
+
+
+def _require_ai_enabled() -> None:
+    """Raise 503 if AI is not configured."""
+    if not settings.ai_enabled:
+        raise HTTPException(
+            status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
+            detail="AI fix is not configured. Set GOOGLE_AI_API_KEY or ANTHROPIC_API_KEY.",
+        )
+
+
+@router.post("/fix-tree", response_model=AIFixTreeResponse)
+@limiter.limit("10/minute")
+async def fix_tree(
+    request: Request,
+    body: AIFixTreeRequest,
+    user: Annotated[User, Depends(require_engineer_or_admin)],
+    db: Annotated[AsyncSession, Depends(get_db)],
+):
+    """Generate AI-powered fixes for tree validation errors."""
+    _require_ai_enabled()
+
+    validation_errors = [
+        {"node_id": e.node_id, "message": e.message}
+        for e in body.validation_errors
+    ]
+
+    try:
+        fixes, input_tokens, output_tokens = await generate_fixes(
+            tree_structure=body.tree_structure,
+            tree_name=body.tree_name,
+            tree_type=body.tree_type,
+            validation_errors=validation_errors,
+        )
+    except RuntimeError as exc:
+        logger.error("AI provider not available: %s", exc)
+        raise HTTPException(
+            status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
+            detail=str(exc),
+        )
+    except Exception as exc:
+        logger.exception("Unexpected error in AI fix service")
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail="An unexpected error occurred while generating fixes.",
+        )
+
+    return AIFixTreeResponse(
+        fixes=[AIFixProposal(**f) for f in fixes],
+        tokens_used=AIFixTokenUsage(input=input_tokens, output=output_tokens),
+    )
--- a/backend/app/api/router.py
+++ b/backend/app/api/router.py
@@ -6,6 +6,7 @@ from app.api.endpoints import target_lists
 from app.api.endpoints import maintenance_schedules
 from app.api.endpoints import feedback
 from app.api.endpoints import ai_builder
+from app.api.endpoints import ai_fix

 api_router = APIRouter()

@@ -36,3 +37,4 @@ api_router.include_router(target_lists.router)
 api_router.include_router(maintenance_schedules.router)
 api_router.include_router(feedback.router)
 api_router.include_router(ai_builder.router)
+api_router.include_router(ai_fix.router)
--- a/backend/app/core/ai_fix_service.py
+++ b/backend/app/core/ai_fix_service.py
@@ -0,0 +1,273 @@
+"""AI-powered fix service for tree validation errors.
+
+Given a tree structure and validation errors, generates AI-powered
+proposals to fix each structural issue while preserving existing content.
+"""
+
+import copy
+import json
+import logging
+import re
+from typing import Any
+
+from app.core.ai_provider import get_ai_provider
+from app.core.ai_tree_validator import validate_generated_tree
+
+logger = logging.getLogger(__name__)
+
+
+FIX_SYSTEM_PROMPT = """You are ResolutionFlow AI, fixing structural validation errors in IT troubleshooting and maintenance flows used by MSP engineers.
+
+You will receive:
+1. A full flow outline showing the tree structure
+2. The specific failing node with its full JSON
+3. The validation error message
+
+Your task: Return a FIXED version of the failing node as valid JSON. Rules:
+- Fix ONLY the structural issue described in the error message
+- Keep ALL existing content (titles, descriptions, questions, options) unchanged
+- When adding new nodes (e.g., missing branches), generate domain-relevant content that fits the flow's topic
+- Every new node must have a unique ID (use descriptive kebab-case IDs)
+- Decision nodes must have at least 2 options and at least 2 children
+- Action nodes must have a next_node_id pointing to a sibling node in the parent's children
+- Solution nodes are leaf nodes (no children)
+- Return ONLY the fixed node JSON, no explanation"""
+
+
+# ── Pure helper functions ──
+
+
+def _find_node_by_id(tree: dict[str, Any], node_id: str) -> dict[str, Any] | None:
+    """Recursively find a node by its ID in the tree structure."""
+    if not isinstance(tree, dict):
+        return None
+    if tree.get("id") == node_id:
+        return tree
+    for child in tree.get("children", []):
+        result = _find_node_by_id(child, node_id)
+        if result is not None:
+            return result
+    return None
+
+
+def _find_parent_node(tree: dict[str, Any], target_id: str) -> dict[str, Any] | None:
+    """Find the parent of a node with the given ID."""
+    if not isinstance(tree, dict):
+        return None
+    for child in tree.get("children", []):
+        if isinstance(child, dict) and child.get("id") == target_id:
+            return tree
+        result = _find_parent_node(child, target_id)
+        if result is not None:
+            return result
+    return None
+
+
+def _serialize_tree_outline(
+    tree: dict[str, Any], indent: int = 0, error_node_id: str | None = None
+) -> str:
+    """Serialize tree as a readable outline for AI prompt context.
+
+    Format: indented "- [type] label" with "<<< ERROR HERE" marker.
+    """
+    if not isinstance(tree, dict):
+        return ""
+
+    node_type = tree.get("type", "unknown")
+    label = tree.get("question") or tree.get("title") or tree.get("id", "?")
+    prefix = "  " * indent
+    marker = " <<< ERROR HERE" if tree.get("id") == error_node_id else ""
+    line = f"{prefix}- [{node_type}] {label}{marker}"
+
+    lines = [line]
+    for child in tree.get("children", []):
+        lines.append(_serialize_tree_outline(child, indent + 1, error_node_id))
+
+    return "\n".join(lines)
+
+
+def _strip_markdown_fences(text: str) -> str:
+    """Strip ```json...``` fences from AI response."""
+    return re.sub(r"^```(?:json)?\s*\n?", "", text.strip(), flags=re.MULTILINE).rstrip(
+        "`"
+    ).strip()
+
+
+def _replace_node_in_tree(
+    tree: dict[str, Any], target_id: str, replacement: dict[str, Any]
+) -> bool:
+    """Replace a node in-place by ID. Returns True if found and replaced."""
+    if not isinstance(tree, dict):
+        return False
+    if tree.get("id") == target_id:
+        tree.clear()
+        tree.update(replacement)
+        return True
+    for child in tree.get("children", []):
+        if _replace_node_in_tree(child, target_id, replacement):
+            return True
+    return False
+
+
+def _describe_fix(original: dict[str, Any], fixed: dict[str, Any]) -> str:
+    """Describe what changed between original and fixed node."""
+    changes: list[str] = []
+
+    orig_children = len(original.get("children", []))
+    fixed_children = len(fixed.get("children", []))
+    if fixed_children > orig_children:
+        changes.append(f"added {fixed_children - orig_children} child node(s)")
+
+    orig_options = len(original.get("options", []))
+    fixed_options = len(fixed.get("options", []))
+    if fixed_options > orig_options:
+        changes.append(f"added {fixed_options - orig_options} option(s)")
+
+    if fixed.get("next_node_id") and not original.get("next_node_id"):
+        changes.append("added next_node_id")
+
+    if not changes:
+        changes.append("fixed structural issue")
+
+    return "; ".join(changes).capitalize()
+
+
+# ── Prompt building ──
+
+
+def _build_fix_prompt(
+    tree: dict[str, Any],
+    node_id: str,
+    error_message: str,
+    tree_name: str,
+    tree_type: str,
+) -> str:
+    """Build the user message for the AI fix request."""
+    outline = _serialize_tree_outline(tree, error_node_id=node_id)
+    node = _find_node_by_id(tree, node_id)
+    node_json = json.dumps(node, indent=2) if node else "{}"
+
+    return (
+        f"Flow name: {tree_name}\n"
+        f"Flow type: {tree_type}\n\n"
+        f"## Full flow outline\n```\n{outline}\n```\n\n"
+        f"## Failing node (ID: {node_id})\n```json\n{node_json}\n```\n\n"
+        f"## Validation error\n{error_message}\n\n"
+        f"Return the fixed version of this node as JSON."
+    )
+
+
+# ── Main entry point ──
+
+
+async def generate_fixes(
+    tree_structure: dict[str, Any],
+    tree_name: str,
+    tree_type: str,
+    validation_errors: list[dict[str, str]],
+) -> tuple[list[dict[str, Any]], int, int]:
+    """Generate AI-powered fixes for tree validation errors.
+
+    Args:
+        tree_structure: Full tree structure dict.
+        tree_name: Name of the flow.
+        tree_type: Type of flow (troubleshooting, procedural, maintenance).
+        validation_errors: List of dicts with "node_id" and "message" keys.
+
+    Returns:
+        Tuple of (fixes_list, total_input_tokens, total_output_tokens).
+        Each fix dict has: target_node_id, error_message, description,
+        original_node, fixed_node.
+    """
+    provider = get_ai_provider()
+    fixes: list[dict[str, Any]] = []
+    total_input_tokens = 0
+    total_output_tokens = 0
+
+    for error in validation_errors:
+        node_id = error["node_id"]
+        error_message = error["message"]
+
+        original_node = _find_node_by_id(tree_structure, node_id)
+        if original_node is None:
+            logger.warning("Node %s not found in tree, skipping fix", node_id)
+            continue
+
+        original_snapshot = copy.deepcopy(original_node)
+
+        # Build prompt and call AI
+        user_message = _build_fix_prompt(
+            tree_structure, node_id, error_message, tree_name, tree_type
+        )
+        messages = [{"role": "user", "content": user_message}]
+
+        try:
+            text, in_tok, out_tok = await provider.generate_json(
+                system_prompt=FIX_SYSTEM_PROMPT,
+                messages=messages,
+                max_tokens=2048,
+            )
+            total_input_tokens += in_tok
+            total_output_tokens += out_tok
+
+            cleaned = _strip_markdown_fences(text)
+            fixed_node = json.loads(cleaned)
+        except (json.JSONDecodeError, Exception) as exc:
+            logger.warning("AI fix failed for node %s: %s", node_id, exc)
+            continue
+
+        # Validate by substituting into a tree copy
+        tree_copy = copy.deepcopy(tree_structure)
+        _replace_node_in_tree(tree_copy, node_id, copy.deepcopy(fixed_node))
+        remaining_errors = validate_generated_tree(tree_copy)
+
+        # Check if the specific error is still present
+        still_has_error = any(node_id in e for e in remaining_errors)
+
+        if still_has_error:
+            # Retry once with corrective prompt
+            retry_message = (
+                f"Your previous fix still has validation errors:\n"
+                f"{chr(10).join(remaining_errors)}\n\n"
+                f"Please fix the node again. Return ONLY the corrected JSON."
+            )
+            messages.append({"role": "assistant", "content": text})
+            messages.append({"role": "user", "content": retry_message})
+
+            try:
+                text2, in_tok2, out_tok2 = await provider.generate_json(
+                    system_prompt=FIX_SYSTEM_PROMPT,
+                    messages=messages,
+                    max_tokens=2048,
+                )
+                total_input_tokens += in_tok2
+                total_output_tokens += out_tok2
+
+                cleaned2 = _strip_markdown_fences(text2)
+                fixed_node = json.loads(cleaned2)
+
+                # Re-validate
+                tree_copy2 = copy.deepcopy(tree_structure)
+                _replace_node_in_tree(tree_copy2, node_id, copy.deepcopy(fixed_node))
+                remaining2 = validate_generated_tree(tree_copy2)
+                still_has_error = any(node_id in e for e in remaining2)
+            except (json.JSONDecodeError, Exception) as exc:
+                logger.warning("AI retry fix failed for node %s: %s", node_id, exc)
+                continue
+
+        if still_has_error:
+            logger.warning("AI could not fix node %s after retry", node_id)
+            continue
+
+        description = _describe_fix(original_snapshot, fixed_node)
+        fixes.append(
+            {
+                "target_node_id": node_id,
+                "error_message": error_message,
+                "description": description,
+                "original_node": original_snapshot,
+                "fixed_node": fixed_node,
+            }
+        )
+
+    return fixes, total_input_tokens, total_output_tokens
--- a/backend/app/core/ai_provider.py
+++ b/backend/app/core/ai_provider.py
@@ -0,0 +1,175 @@
+"""
+AI Provider abstraction layer.
+
+Supports Gemini (google-genai) and Anthropic (anthropic) as interchangeable
+backends for JSON generation used by the AI Flow Builder.
+"""
+
+import logging
+from abc import ABC, abstractmethod
+
+from app.core.config import settings
+
+logger = logging.getLogger(__name__)
+
+
+class AIProvider(ABC):
+    """Abstract base class for AI providers."""
+
+    @abstractmethod
+    async def generate_json(
+        self,
+        system_prompt: str,
+        messages: list[dict[str, str]],
+        max_tokens: int = 4096,
+    ) -> tuple[str, int, int]:
+        """Generate a JSON response from the AI model.
+
+        Args:
+            system_prompt: System-level instruction for the model.
+            messages: List of message dicts with "role" and "content" keys.
+            max_tokens: Maximum output tokens.
+
+        Returns:
+            Tuple of (response_text, input_tokens, output_tokens).
+        """
+        ...
+
+
+class GeminiProvider(AIProvider):
+    """Google Gemini provider using the google-genai SDK."""
+
+    def __init__(self, api_key: str, model: str) -> None:
+        self._api_key = api_key
+        self._model = model
+
+    async def generate_json(
+        self,
+        system_prompt: str,
+        messages: list[dict[str, str]],
+        max_tokens: int = 4096,
+    ) -> tuple[str, int, int]:
+        from google import genai
+        from google.genai import types as genai_types
+
+        client = genai.Client(api_key=self._api_key)
+
+        # Convert messages to Gemini Content format
+        contents: list[genai_types.Content] = []
+        for msg in messages:
+            role = "model" if msg["role"] == "assistant" else "user"
+            contents.append(
+                genai_types.Content(
+                    role=role,
+                    parts=[genai_types.Part(text=msg["content"])],
+                )
+            )
+
+        config = genai_types.GenerateContentConfig(
+            system_instruction=system_prompt,
+            max_output_tokens=max_tokens,
+            response_mime_type="application/json",
+        )
+
+        response = await client.aio.models.generate_content(
+            model=self._model,
+            contents=contents,
+            config=config,
+        )
+
+        # Log finish reason to detect truncation
+        if response.candidates:
+            finish_reason = getattr(response.candidates[0], "finish_reason", None)
+            logger.info("Gemini finish_reason=%s model=%s", finish_reason, self._model)
+            if str(finish_reason) == "MAX_TOKENS":
+                logger.warning(
+                    "Gemini output truncated (MAX_TOKENS). max_output_tokens=%d",
+                    max_tokens,
+                )
+
+        text = response.text or ""
+        input_tokens = getattr(response.usage_metadata, "prompt_token_count", 0) or 0
+        output_tokens = (
+            getattr(response.usage_metadata, "candidates_token_count", 0) or 0
+        )
+
+        return text, input_tokens, output_tokens
+
+
+class AnthropicProvider(AIProvider):
+    """Anthropic Claude provider using the anthropic SDK."""
+
+    def __init__(self, api_key: str, model: str, timeout: int = 45) -> None:
+        self._api_key = api_key
+        self._model = model
+        self._timeout = timeout
+
+    async def generate_json(
+        self,
+        system_prompt: str,
+        messages: list[dict[str, str]],
+        max_tokens: int = 4096,
+    ) -> tuple[str, int, int]:
+        import anthropic
+
+        client = anthropic.AsyncAnthropic(
+            api_key=self._api_key,
+            timeout=self._timeout,
+        )
+
+        response = await client.messages.create(
+            model=self._model,
+            max_tokens=max_tokens,
+            system=system_prompt,
+            messages=messages,
+        )
+
+        text = response.content[0].text
+        input_tokens = response.usage.input_tokens
+        output_tokens = response.usage.output_tokens
+
+        return text, input_tokens, output_tokens
+
+
+def get_ai_provider() -> AIProvider:
+    """Factory that returns the configured AI provider.
+
+    Selection logic:
+    1. If AI_PROVIDER == "gemini" and GOOGLE_AI_API_KEY is set -> GeminiProvider
+    2. If AI_PROVIDER == "anthropic" and ANTHROPIC_API_KEY is set -> AnthropicProvider
+    3. Fallback: if preferred provider key missing, try the other one
+    4. If nothing configured -> raise RuntimeError
+    """
+    provider = settings.AI_PROVIDER
+
+    if provider == "gemini":
+        if settings.GOOGLE_AI_API_KEY:
+            return GeminiProvider(
+                api_key=settings.GOOGLE_AI_API_KEY,
+                model=settings.AI_MODEL_GEMINI,
+            )
+        # Fallback to Anthropic
+        if settings.ANTHROPIC_API_KEY:
+            return AnthropicProvider(
+                api_key=settings.ANTHROPIC_API_KEY,
+                model=settings.AI_MODEL_ANTHROPIC,
+                timeout=settings.AI_REQUEST_TIMEOUT_SECONDS,
+            )
+
+    elif provider == "anthropic":
+        if settings.ANTHROPIC_API_KEY:
+            return AnthropicProvider(
+                api_key=settings.ANTHROPIC_API_KEY,
+                model=settings.AI_MODEL_ANTHROPIC,
+                timeout=settings.AI_REQUEST_TIMEOUT_SECONDS,
+            )
+        # Fallback to Gemini
+        if settings.GOOGLE_AI_API_KEY:
+            return GeminiProvider(
+                api_key=settings.GOOGLE_AI_API_KEY,
+                model=settings.AI_MODEL_GEMINI,
+            )
+
+    raise RuntimeError(
+        "No AI provider configured. Set GOOGLE_AI_API_KEY or ANTHROPIC_API_KEY."
+    )
--- a/backend/app/core/ai_tree_generator_service.py
+++ b/backend/app/core/ai_tree_generator_service.py
@@ -1,11 +1,11 @@
-"""AI-powered tree generation service using Anthropic Claude API.
+"""AI-powered tree generation service.

 Implements the 4-stage wizard flow:
  Stage 2 (scaffold): AI suggests 4-7 top-level branches
  Stage 3 (branch_detail): AI generates detailed nodes per branch
  Stage 4 (assemble): Pure assembly logic — zero AI calls

-System prompts are static constants to enable Anthropic prompt caching.
+Uses the provider abstraction from ai_provider.py (supports Gemini + Anthropic).
 """
 import json
 import logging
@@ -13,8 +13,7 @@ import re
 import uuid
 from typing import Any

-import anthropic
-
+from app.core.ai_provider import get_ai_provider
 from app.core.config import settings
 from app.core.ai_tree_validator import validate_generated_tree, count_tree_stats

@@ -121,15 +120,6 @@ def _strip_markdown_fences(text: str) -> str:
    return text


-def _get_client() -> anthropic.AsyncAnthropic:
-    """Get configured async Anthropic client."""
-    if not settings.ANTHROPIC_API_KEY:
-        raise RuntimeError("ANTHROPIC_API_KEY not configured")
-    return anthropic.AsyncAnthropic(
-        api_key=settings.ANTHROPIC_API_KEY,
-        timeout=settings.AI_REQUEST_TIMEOUT_SECONDS,
-    )
-

 def _estimate_cost(input_tokens: int, output_tokens: int) -> float:
    """Estimate USD cost from token counts."""
@@ -146,7 +136,7 @@ async def scaffold_branches(
    Returns (branches, input_tokens, output_tokens, estimated_cost).
    Raises ValueError on invalid response.
    """
-    client = _get_client()
+    provider = get_ai_provider()

    flow_type = wizard_state.get("flow_type", "troubleshooting")
    name = wizard_state.get("name", "")
@@ -161,21 +151,26 @@ async def scaffold_branches(
    if tags:
        user_message += f"Environment: {', '.join(tags)}\n"

-    response = await client.messages.create(
-        model=settings.AI_MODEL,
-        max_tokens=1024,
-        system=SCAFFOLD_SYSTEM_PROMPT,
+    raw_text, input_tokens, output_tokens = await provider.generate_json(
+        system_prompt=SCAFFOLD_SYSTEM_PROMPT,
        messages=[{"role": "user", "content": user_message}],
+        max_tokens=2048,
    )

-    raw_text = _strip_markdown_fences(response.content[0].text)
-    input_tokens = response.usage.input_tokens
-    output_tokens = response.usage.output_tokens
+    logger.info(
+        "scaffold raw response (tokens in=%d out=%d, len=%d): %s",
+        input_tokens,
+        output_tokens,
+        len(raw_text),
+        raw_text[:500],
+    )
+    raw_text = _strip_markdown_fences(raw_text)
    cost = _estimate_cost(input_tokens, output_tokens)

    try:
        data = json.loads(raw_text)
    except json.JSONDecodeError as e:
+        logger.error("scaffold JSON parse failed. Full text (%d chars): %s", len(raw_text), raw_text)
        raise ValueError(f"AI returned invalid JSON: {e}")

    branches = data.get("branches", [])
@@ -196,7 +191,7 @@ async def generate_branch_detail(
    On validation failure, retries once with corrective prompt.
    Raises ValueError if both attempts fail.
    """
-    client = _get_client()
+    provider = get_ai_provider()

    flow_type = wizard_state.get("flow_type", "troubleshooting")
    name = wizard_state.get("name", "")
@@ -217,35 +212,30 @@ async def generate_branch_detail(
    total_output = 0

    for attempt in range(3):
-        response = await client.messages.create(
-            model=settings.AI_MODEL,
-            max_tokens=8192,
-            system=BRANCH_DETAIL_SYSTEM_PROMPT,
+        raw_text, input_tokens, output_tokens = await provider.generate_json(
+            system_prompt=BRANCH_DETAIL_SYSTEM_PROMPT,
            messages=messages,
+            max_tokens=8192,
        )

-        total_input += response.usage.input_tokens
-        total_output += response.usage.output_tokens
+        total_input += input_tokens
+        total_output += output_tokens
        logger.debug(
-            "branch_detail attempt=%d stop_reason=%s content_blocks=%d output_tokens=%d",
+            "branch_detail attempt=%d output_tokens=%d",
            attempt,
-            response.stop_reason,
-            len(response.content),
-            response.usage.output_tokens,
+            output_tokens,
        )
-        if response.stop_reason == "max_tokens":
-            logger.warning(
-                "branch_detail attempt=%d hit max_tokens limit (%d output tokens) — response may be truncated",
-                attempt,
-                response.usage.output_tokens,
-            )
-        raw_text = _strip_markdown_fences(response.content[0].text) if response.content else ""
+        raw_text = _strip_markdown_fences(raw_text) if raw_text else ""
        if not raw_text:
-            logger.warning("branch_detail attempt=%d returned empty text, stop_reason=%s", attempt, response.stop_reason)
+            logger.warning("branch_detail attempt=%d returned empty text", attempt)

        try:
            branch_tree = json.loads(raw_text)
        except json.JSONDecodeError as e:
+            logger.error(
+                "branch_detail attempt=%d JSON parse failed (%d chars): %s",
+                attempt, len(raw_text), raw_text[:500],
+            )
            if attempt < 2:
                messages.append({"role": "assistant", "content": raw_text})
                messages.append({
--- a/backend/app/core/config.py
+++ b/backend/app/core/config.py
@@ -78,11 +78,16 @@ class Settings(BaseSettings):
    AI_CONVERSATION_TTL_HOURS: int = 24
    AI_MAX_CALLS_PER_FLOW: int = 10
    AI_REQUEST_TIMEOUT_SECONDS: int = 45
+    # AI Provider selection
+    AI_PROVIDER: str = "gemini"  # "gemini" or "anthropic"
+    GOOGLE_AI_API_KEY: Optional[str] = None
+    AI_MODEL_GEMINI: str = "gemini-2.5-flash"
+    AI_MODEL_ANTHROPIC: str = "claude-haiku-4-5-20251001"

    @property
    def ai_enabled(self) -> bool:
-        """Check if AI Flow Builder is configured."""
-        return self.ANTHROPIC_API_KEY is not None
+        """Check if any AI provider is configured."""
+        return self.ANTHROPIC_API_KEY is not None or self.GOOGLE_AI_API_KEY is not None

    # Deployment – auto-seed test data on PR environments
    SEED_ON_DEPLOY: bool = False
--- a/backend/app/schemas/ai_fix.py
+++ b/backend/app/schemas/ai_fix.py
@@ -0,0 +1,52 @@
+"""Pydantic schemas for the AI auto-fix feature."""
+from typing import Any, Literal
+
+from pydantic import BaseModel, Field
+
+
+# ── Requests ──
+
+
+class ValidationErrorInput(BaseModel):
+    """A single validation error to fix."""
+
+    node_id: str = Field(..., description="ID of the node with the error")
+    message: str = Field(..., description="The validation error message")
+
+
+class AIFixTreeRequest(BaseModel):
+    """Request to generate AI fixes for validation errors."""
+
+    tree_structure: dict[str, Any] = Field(..., description="Full tree structure")
+    tree_name: str = Field("", max_length=255, description="Name of the flow")
+    tree_type: Literal["troubleshooting", "procedural", "maintenance"] = Field(
+        "troubleshooting", description="Type of flow"
+    )
+    validation_errors: list[ValidationErrorInput] = Field(
+        ..., min_length=1, max_length=10, description="Errors to fix"
+    )
+
+
+# ── Responses ──
+
+
+class AIFixProposal(BaseModel):
+    """A single proposed fix from the AI."""
+
+    target_node_id: str
+    error_message: str
+    description: str
+    original_node: dict[str, Any]
+    fixed_node: dict[str, Any]
+
+
+class AIFixTokenUsage(BaseModel):
+    input: int = 0
+    output: int = 0
+
+
+class AIFixTreeResponse(BaseModel):
+    """Response with proposed fixes."""
+
+    fixes: list[AIFixProposal]
+    tokens_used: AIFixTokenUsage
--- a/backend/requirements.txt
+++ b/backend/requirements.txt
@@ -33,6 +33,7 @@ httpx>=0.27.0

 # AI Flow Builder
 anthropic>=0.40.0
+google-genai>=1.0.0

 # Utilities
 python-dotenv==1.0.1
--- a/backend/tests/test_ai_endpoints.py
+++ b/backend/tests/test_ai_endpoints.py
@@ -1,6 +1,6 @@
 """Integration tests for AI Flow Builder endpoints.

-All Anthropic API calls are mocked — zero real API spend.
+All AI provider calls are mocked — zero real API spend.
 """
 import json
 from unittest.mock import AsyncMock, patch, MagicMock
@@ -64,12 +64,11 @@ BRANCH_DETAIL_JSON = json.dumps({
 })


-def _mock_anthropic_response(text: str, input_tokens: int = 100, output_tokens: int = 200):
-    """Create a mock Anthropic API response."""
-    response = MagicMock()
-    response.content = [MagicMock(text=text)]
-    response.usage = MagicMock(input_tokens=input_tokens, output_tokens=output_tokens)
-    return response
+def _mock_ai_provider(text: str, input_tokens: int = 100, output_tokens: int = 200):
+    """Create a mock AI provider whose generate_json returns the given text and token counts."""
+    provider = MagicMock()
+    provider.generate_json = AsyncMock(return_value=(text, input_tokens, output_tokens))
+    return provider


@pytest.fixture
@@ -194,11 +193,9 @@ async def test_scaffold_success(client, auth_headers, enable_ai):
    )
    conversation_id = start_resp.json()["conversation_id"]

-    # Mock Anthropic
-    mock_response = _mock_anthropic_response(SCAFFOLD_RESPONSE_JSON)
-    with patch("app.core.ai_tree_generator_service._get_client") as mock_client:
-        mock_client.return_value.messages.create = AsyncMock(return_value=mock_response)
-
+    # Mock AI provider
+    mock_provider = _mock_ai_provider(SCAFFOLD_RESPONSE_JSON)
+    with patch("app.core.ai_tree_generator_service.get_ai_provider", return_value=mock_provider):
        response = await client.post(
            "/api/v1/ai/scaffold",
            json={"conversation_id": conversation_id},
@@ -241,9 +238,8 @@ async def test_branch_detail_success(client, auth_headers, enable_ai):
    )
    conversation_id = start_resp.json()["conversation_id"]

-    scaffold_mock = _mock_anthropic_response(SCAFFOLD_RESPONSE_JSON)
-    with patch("app.core.ai_tree_generator_service._get_client") as mock_client:
-        mock_client.return_value.messages.create = AsyncMock(return_value=scaffold_mock)
+    scaffold_provider = _mock_ai_provider(SCAFFOLD_RESPONSE_JSON)
+    with patch("app.core.ai_tree_generator_service.get_ai_provider", return_value=scaffold_provider):
        await client.post(
            "/api/v1/ai/scaffold",
            json={"conversation_id": conversation_id},
@@ -251,10 +247,8 @@ async def test_branch_detail_success(client, auth_headers, enable_ai):
        )

    # Now generate branch detail
-    detail_mock = _mock_anthropic_response(BRANCH_DETAIL_JSON)
-    with patch("app.core.ai_tree_generator_service._get_client") as mock_client:
-        mock_client.return_value.messages.create = AsyncMock(return_value=detail_mock)
-
+    detail_provider = _mock_ai_provider(BRANCH_DETAIL_JSON)
+    with patch("app.core.ai_tree_generator_service.get_ai_provider", return_value=detail_provider):
        response = await client.post(
            "/api/v1/ai/branch-detail",
            json={
@@ -290,9 +284,8 @@ async def test_assemble_success(client, auth_headers, enable_ai):
    conversation_id = start_resp.json()["conversation_id"]

    # Scaffold
-    scaffold_mock = _mock_anthropic_response(SCAFFOLD_RESPONSE_JSON)
-    with patch("app.core.ai_tree_generator_service._get_client") as mock_client:
-        mock_client.return_value.messages.create = AsyncMock(return_value=scaffold_mock)
+    scaffold_provider = _mock_ai_provider(SCAFFOLD_RESPONSE_JSON)
+    with patch("app.core.ai_tree_generator_service.get_ai_provider", return_value=scaffold_provider):
        await client.post(
            "/api/v1/ai/scaffold",
            json={"conversation_id": conversation_id},
--- a/backend/tests/test_ai_fix_endpoint.py
+++ b/backend/tests/test_ai_fix_endpoint.py
@@ -0,0 +1,169 @@
+"""Integration tests for the POST /ai/fix-tree endpoint."""
+
+import json
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+
+from app.core.config import settings
+
+
+# ── Sample tree (has a decision node with only 1 option + 1 child) ──
+
+SAMPLE_TREE = {
+    "id": "root",
+    "type": "decision",
+    "question": "Is the server up?",
+    "options": [
+        {"id": "opt-yes", "label": "Yes", "next_node_id": "check-logs"},
+        {"id": "opt-no", "label": "No", "next_node_id": "restart"},
+    ],
+    "children": [
+        {
+            "id": "check-logs",
+            "type": "action",
+            "title": "Check Logs",
+            "description": "Review logs.",
+            "next_node_id": "logs-ok",
+        },
+        {
+            "id": "logs-ok",
+            "type": "solution",
+            "title": "Logs OK",
+            "description": "Issue in logs.",
+        },
+        {
+            "id": "restart",
+            "type": "decision",
+            "question": "Did restart work?",
+            "options": [{"id": "opt-r", "label": "Yes", "next_node_id": "done"}],
+            "children": [
+                {
+                    "id": "done",
+                    "type": "solution",
+                    "title": "Done",
+                    "description": "Fixed.",
+                }
+            ],
+        },
+    ],
+}
+
+# Fixed version of the "restart" node — 2 options, 2 children
+FIXED_RESTART_NODE = {
+    "id": "restart",
+    "type": "decision",
+    "question": "Did restart work?",
+    "options": [
+        {"id": "opt-r-yes", "label": "Yes", "next_node_id": "done"},
+        {"id": "opt-r-no", "label": "No", "next_node_id": "escalate"},
+    ],
+    "children": [
+        {
+            "id": "done",
+            "type": "solution",
+            "title": "Done",
+            "description": "Fixed.",
+        },
+        {
+            "id": "escalate",
+            "type": "solution",
+            "title": "Escalate",
+            "description": "Escalate to senior engineer.",
+        },
+    ],
+}
+
+FIX_REQUEST_BODY = {
+    "tree_structure": SAMPLE_TREE,
+    "tree_name": "Server Troubleshooting",
+    "tree_type": "troubleshooting",
+    "validation_errors": [
+        {
+            "node_id": "restart",
+            "message": "Decision node 'restart' must have at least 2 options",
+        }
+    ],
+}
+
+
+def _mock_ai_provider(response_text: str, input_tokens: int = 50, output_tokens: int = 100):
+    """Create a mock provider whose generate_json returns given text."""
+    provider = MagicMock()
+    provider.generate_json = AsyncMock(return_value=(response_text, input_tokens, output_tokens))
+    return provider
+
+
+@pytest.fixture
+def enable_ai():
+    """Temporarily enable AI by setting a fake API key."""
+    original = settings.GOOGLE_AI_API_KEY
+    settings.GOOGLE_AI_API_KEY = "test-key-fake"
+    yield
+    settings.GOOGLE_AI_API_KEY = original
+
+
+@pytest.fixture
+def disable_ai():
+    """Ensure AI is disabled."""
+    orig_google = settings.GOOGLE_AI_API_KEY
+    orig_anthropic = settings.ANTHROPIC_API_KEY
+    settings.GOOGLE_AI_API_KEY = None
+    settings.ANTHROPIC_API_KEY = None
+    yield
+    settings.GOOGLE_AI_API_KEY = orig_google
+    settings.ANTHROPIC_API_KEY = orig_anthropic
+
+
+# ── Tests ──
+
+
+@pytest.mark.asyncio
+async def test_returns_401_without_auth(client):
+    """POST /ai/fix-tree without auth token returns 401."""
+    response = await client.post("/api/v1/ai/fix-tree", json=FIX_REQUEST_BODY)
+    assert response.status_code == 401
+
+
+@pytest.mark.asyncio
+async def test_returns_503_when_ai_disabled(client, auth_headers, disable_ai):
+    """POST /ai/fix-tree returns 503 when no AI keys are configured."""
+    response = await client.post(
+        "/api/v1/ai/fix-tree",
+        json=FIX_REQUEST_BODY,
+        headers=auth_headers,
+    )
+    assert response.status_code == 503
+
+
+@pytest.mark.asyncio
+async def test_returns_fixes_on_success(client, auth_headers, enable_ai):
+    """POST /ai/fix-tree returns fix proposals when AI succeeds."""
+    mock_provider = _mock_ai_provider(json.dumps(FIXED_RESTART_NODE))
+
+    with patch(
+        "app.core.ai_fix_service.get_ai_provider",
+        return_value=mock_provider,
+    ):
+        response = await client.post(
+            "/api/v1/ai/fix-tree",
+            json=FIX_REQUEST_BODY,
+            headers=auth_headers,
+        )
+
+    assert response.status_code == 200
+    data = response.json()
+    assert "fixes" in data
+    assert "tokens_used" in data
+    assert len(data["fixes"]) == 1
+
+    fix = data["fixes"][0]
+    assert fix["target_node_id"] == "restart"
+    assert fix["error_message"] == "Decision node 'restart' must have at least 2 options"
+    assert fix["original_node"]["id"] == "restart"
+    assert fix["fixed_node"]["id"] == "restart"
+    assert len(fix["fixed_node"]["options"]) == 2
+    assert len(fix["fixed_node"]["children"]) == 2
+
+    assert data["tokens_used"]["input"] == 50
+    assert data["tokens_used"]["output"] == 100
--- a/backend/tests/test_ai_fix_service.py
+++ b/backend/tests/test_ai_fix_service.py
@@ -0,0 +1,224 @@
+"""Unit tests for AI fix service helper functions.
+
+Tests pure Python helpers only — no AI mocking needed.
+"""
+
+import pytest
+
+from app.core.ai_fix_service import (
+    _find_node_by_id,
+    _find_parent_node,
+    _serialize_tree_outline,
+    _strip_markdown_fences,
+    _replace_node_in_tree,
+    _describe_fix,
+)
+
+
+# ── Sample tree ──
+
+SAMPLE_TREE = {
+    "id": "root",
+    "type": "decision",
+    "question": "Is the server up?",
+    "options": [
+        {"id": "opt-yes", "label": "Yes", "next_node_id": "check-logs"},
+        {"id": "opt-no", "label": "No", "next_node_id": "restart"},
+    ],
+    "children": [
+        {
+            "id": "check-logs",
+            "type": "action",
+            "title": "Check Logs",
+            "description": "Review logs.",
+            "next_node_id": "logs-ok",
+        },
+        {
+            "id": "logs-ok",
+            "type": "solution",
+            "title": "Logs OK",
+            "description": "Issue in logs.",
+        },
+        {
+            "id": "restart",
+            "type": "decision",
+            "question": "Did restart work?",
+            "options": [{"id": "opt-r", "label": "Yes", "next_node_id": "done"}],
+            "children": [
+                {
+                    "id": "done",
+                    "type": "solution",
+                    "title": "Done",
+                    "description": "Fixed.",
+                }
+            ],
+        },
+    ],
+}
+
+
+# ── _find_node_by_id ──
+
+
+class TestFindNodeById:
+    def test_finds_root(self):
+        node = _find_node_by_id(SAMPLE_TREE, "root")
+        assert node is not None
+        assert node["id"] == "root"
+        assert node["type"] == "decision"
+
+    def test_finds_nested_child(self):
+        node = _find_node_by_id(SAMPLE_TREE, "done")
+        assert node is not None
+        assert node["id"] == "done"
+        assert node["type"] == "solution"
+
+    def test_finds_direct_child(self):
+        node = _find_node_by_id(SAMPLE_TREE, "check-logs")
+        assert node is not None
+        assert node["title"] == "Check Logs"
+
+    def test_returns_none_for_missing(self):
+        node = _find_node_by_id(SAMPLE_TREE, "nonexistent")
+        assert node is None
+
+    def test_returns_none_for_non_dict(self):
+        assert _find_node_by_id("not a dict", "root") is None
+
+
+# ── _find_parent_node ──
+
+
+class TestFindParentNode:
+    def test_root_has_no_parent(self):
+        parent = _find_parent_node(SAMPLE_TREE, "root")
+        assert parent is None
+
+    def test_finds_parent_of_direct_child(self):
+        parent = _find_parent_node(SAMPLE_TREE, "check-logs")
+        assert parent is not None
+        assert parent["id"] == "root"
+
+    def test_finds_parent_of_deeply_nested(self):
+        parent = _find_parent_node(SAMPLE_TREE, "done")
+        assert parent is not None
+        assert parent["id"] == "restart"
+
+    def test_returns_none_for_missing(self):
+        parent = _find_parent_node(SAMPLE_TREE, "nonexistent")
+        assert parent is None
+
+    def test_returns_none_for_non_dict(self):
+        assert _find_parent_node("not a dict", "root") is None
+
+
+# ── _serialize_tree_outline ──
+
+
+class TestSerializeTreeOutline:
+    def test_produces_readable_outline(self):
+        outline = _serialize_tree_outline(SAMPLE_TREE)
+        assert "- [decision] Is the server up?" in outline
+        assert "  - [action] Check Logs" in outline
+        assert "  - [solution] Logs OK" in outline
+        assert "    - [solution] Done" in outline
+
+    def test_marks_error_node(self):
+        outline = _serialize_tree_outline(SAMPLE_TREE, error_node_id="restart")
+        assert "<<< ERROR HERE" in outline
+        # Only the restart node should be marked
+        lines = outline.split("\n")
+        error_lines = [l for l in lines if "ERROR HERE" in l]
+        assert len(error_lines) == 1
+        assert "Did restart work?" in error_lines[0]
+
+    def test_no_error_marker_when_none(self):
+        outline = _serialize_tree_outline(SAMPLE_TREE)
+        assert "ERROR HERE" not in outline
+
+    def test_handles_non_dict(self):
+        assert _serialize_tree_outline("not a dict") == ""
+
+    def test_indentation_increases_with_depth(self):
+        outline = _serialize_tree_outline(SAMPLE_TREE)
+        lines = outline.split("\n")
+        # Root has no indentation
+        assert lines[0].startswith("- [decision]")
+        # Children have 2-space indent
+        child_lines = [l for l in lines if "Check Logs" in l]
+        assert child_lines[0].startswith("  - ")
+
+
+# ── _strip_markdown_fences ──
+
+
+class TestStripMarkdownFences:
+    def test_strips_json_fences(self):
+        text = '```json\n{"key": "value"}\n```'
+        assert _strip_markdown_fences(text) == '{"key": "value"}'
+
+    def test_strips_plain_fences(self):
+        text = '```\n{"key": "value"}\n```'
+        assert _strip_markdown_fences(text) == '{"key": "value"}'
+
+    def test_passes_through_plain_json(self):
+        text = '{"key": "value"}'
+        assert _strip_markdown_fences(text) == '{"key": "value"}'
+
+
+# ── _replace_node_in_tree ──
+
+
+class TestReplaceNodeInTree:
+    def test_replaces_root(self):
+        import copy
+
+        tree = copy.deepcopy(SAMPLE_TREE)
+        replacement = {"id": "root", "type": "decision", "question": "New question"}
+        assert _replace_node_in_tree(tree, "root", replacement) is True
+        assert tree["question"] == "New question"
+        assert "children" not in tree  # cleared and replaced
+
+    def test_replaces_nested_node(self):
+        import copy
+
+        tree = copy.deepcopy(SAMPLE_TREE)
+        replacement = {"id": "done", "type": "solution", "title": "All Done", "description": "Complete."}
+        assert _replace_node_in_tree(tree, "done", replacement) is True
+        found = _find_node_by_id(tree, "done")
+        assert found["title"] == "All Done"
+
+    def test_returns_false_for_missing(self):
+        import copy
+
+        tree = copy.deepcopy(SAMPLE_TREE)
+        assert _replace_node_in_tree(tree, "nonexistent", {"id": "x"}) is False
+
+
+# ── _describe_fix ──
+
+
+class TestDescribeFix:
+    def test_describes_added_children(self):
+        original = {"id": "n1", "children": [{"id": "c1"}]}
+        fixed = {"id": "n1", "children": [{"id": "c1"}, {"id": "c2"}]}
+        desc = _describe_fix(original, fixed)
+        assert "1 child node" in desc
+
+    def test_describes_added_options(self):
+        original = {"id": "n1", "options": [{"id": "o1"}]}
+        fixed = {"id": "n1", "options": [{"id": "o1"}, {"id": "o2"}]}
+        desc = _describe_fix(original, fixed)
+        assert "1 option" in desc
+
+    def test_describes_added_next_node_id(self):
+        original = {"id": "n1", "type": "action"}
+        fixed = {"id": "n1", "type": "action", "next_node_id": "n2"}
+        desc = _describe_fix(original, fixed)
+        assert "next_node_id" in desc
+
+    def test_fallback_description(self):
+        original = {"id": "n1", "type": "solution"}
+        fixed = {"id": "n1", "type": "solution"}
+        desc = _describe_fix(original, fixed)
+        assert "fixed structural issue" in desc.lower()
--- a/backend/tests/test_ai_provider.py
+++ b/backend/tests/test_ai_provider.py
@@ -0,0 +1,216 @@
+"""Tests for the AI provider abstraction layer."""
+
+import pytest
+from unittest.mock import AsyncMock, MagicMock, patch
+import sys
+
+from app.core.ai_provider import (
+    AIProvider,
+    AnthropicProvider,
+    GeminiProvider,
+    get_ai_provider,
+)
+from app.core.config import settings
+
+
+class TestGetAIProvider:
+    """Tests for the get_ai_provider factory function."""
+
+    def test_returns_gemini_when_configured(self):
+        original_provider = settings.AI_PROVIDER
+        original_key = settings.GOOGLE_AI_API_KEY
+        try:
+            settings.AI_PROVIDER = "gemini"
+            settings.GOOGLE_AI_API_KEY = "test-gemini-key"
+            provider = get_ai_provider()
+            assert isinstance(provider, GeminiProvider)
+        finally:
+            settings.AI_PROVIDER = original_provider
+            settings.GOOGLE_AI_API_KEY = original_key
+
+    def test_returns_anthropic_when_configured(self):
+        original_provider = settings.AI_PROVIDER
+        original_key = settings.ANTHROPIC_API_KEY
+        try:
+            settings.AI_PROVIDER = "anthropic"
+            settings.ANTHROPIC_API_KEY = "test-anthropic-key"
+            provider = get_ai_provider()
+            assert isinstance(provider, AnthropicProvider)
+        finally:
+            settings.AI_PROVIDER = original_provider
+            settings.ANTHROPIC_API_KEY = original_key
+
+    def test_fallback_to_anthropic_when_gemini_key_missing(self):
+        original_provider = settings.AI_PROVIDER
+        original_gemini_key = settings.GOOGLE_AI_API_KEY
+        original_anthropic_key = settings.ANTHROPIC_API_KEY
+        try:
+            settings.AI_PROVIDER = "gemini"
+            settings.GOOGLE_AI_API_KEY = None
+            settings.ANTHROPIC_API_KEY = "test-anthropic-key"
+            provider = get_ai_provider()
+            assert isinstance(provider, AnthropicProvider)
+        finally:
+            settings.AI_PROVIDER = original_provider
+            settings.GOOGLE_AI_API_KEY = original_gemini_key
+            settings.ANTHROPIC_API_KEY = original_anthropic_key
+
+    def test_fallback_to_gemini_when_anthropic_key_missing(self):
+        original_provider = settings.AI_PROVIDER
+        original_gemini_key = settings.GOOGLE_AI_API_KEY
+        original_anthropic_key = settings.ANTHROPIC_API_KEY
+        try:
+            settings.AI_PROVIDER = "anthropic"
+            settings.ANTHROPIC_API_KEY = None
+            settings.GOOGLE_AI_API_KEY = "test-gemini-key"
+            provider = get_ai_provider()
+            assert isinstance(provider, GeminiProvider)
+        finally:
+            settings.AI_PROVIDER = original_provider
+            settings.GOOGLE_AI_API_KEY = original_gemini_key
+            settings.ANTHROPIC_API_KEY = original_anthropic_key
+
+    def test_raises_when_no_provider_configured(self):
+        original_provider = settings.AI_PROVIDER
+        original_gemini_key = settings.GOOGLE_AI_API_KEY
+        original_anthropic_key = settings.ANTHROPIC_API_KEY
+        try:
+            settings.AI_PROVIDER = "gemini"
+            settings.GOOGLE_AI_API_KEY = None
+            settings.ANTHROPIC_API_KEY = None
+            with pytest.raises(RuntimeError, match="No AI provider configured"):
+                get_ai_provider()
+        finally:
+            settings.AI_PROVIDER = original_provider
+            settings.GOOGLE_AI_API_KEY = original_gemini_key
+            settings.ANTHROPIC_API_KEY = original_anthropic_key
+
+
+class TestAnthropicProvider:
+    """Tests for AnthropicProvider.generate_json."""
+
+    @pytest.mark.asyncio
+    async def test_generate_json(self):
+        provider = AnthropicProvider(
+            api_key="test-key", model="claude-haiku-4-5-20251001", timeout=30
+        )
+
+        mock_response = MagicMock()
+        mock_response.content = [MagicMock(text='{"result": "ok"}')]
+        mock_response.usage = MagicMock(input_tokens=100, output_tokens=50)
+
+        mock_client = AsyncMock()
+        mock_client.messages.create = AsyncMock(return_value=mock_response)
+
+        with patch("anthropic.AsyncAnthropic", return_value=mock_client):
+            text, input_tokens, output_tokens = await provider.generate_json(
+                system_prompt="You are a helper.",
+                messages=[{"role": "user", "content": "Hello"}],
+                max_tokens=1024,
+            )
+
+        assert text == '{"result": "ok"}'
+        assert input_tokens == 100
+        assert output_tokens == 50
+
+        mock_client.messages.create.assert_called_once_with(
+            model="claude-haiku-4-5-20251001",
+            max_tokens=1024,
+            system="You are a helper.",
+            messages=[{"role": "user", "content": "Hello"}],
+        )
+
+
+class TestGeminiProvider:
+    """Tests for GeminiProvider.generate_json."""
+
+    @pytest.mark.asyncio
+    async def test_generate_json(self):
+        provider = GeminiProvider(api_key="test-key", model="gemini-2.5-flash")
+
+        mock_usage = MagicMock()
+        mock_usage.prompt_token_count = 80
+        mock_usage.candidates_token_count = 40
+
+        mock_response = MagicMock()
+        mock_response.text = '{"answer": 42}'
+        mock_response.usage_metadata = mock_usage
+
+        mock_client = MagicMock()
+        mock_client.aio.models.generate_content = AsyncMock(
+            return_value=mock_response
+        )
+
+        mock_genai_module = MagicMock()
+        mock_genai_module.Client.return_value = mock_client
+
+        mock_types = MagicMock()
+        mock_types.Content.side_effect = lambda **kw: kw
+        mock_types.Part.side_effect = lambda **kw: kw
+        mock_types.GenerateContentConfig.side_effect = lambda **kw: kw
+
+        mock_google = MagicMock()
+        mock_google.genai = mock_genai_module
+        mock_genai_module.types = mock_types
+
+        with patch.dict(sys.modules, {
+            "google": mock_google,
+            "google.genai": mock_genai_module,
+            "google.genai.types": mock_types,
+        }):
+            text, input_tokens, output_tokens = await provider.generate_json(
+                system_prompt="Generate JSON.",
+                messages=[
+                    {"role": "user", "content": "Give me data"},
+                    {"role": "assistant", "content": "Here it is"},
+                    {"role": "user", "content": "More please"},
+                ],
+                max_tokens=2048,
+            )
+
+        assert text == '{"answer": 42}'
+        assert input_tokens == 80
+        assert output_tokens == 40
+
+        mock_client.aio.models.generate_content.assert_called_once()
+
+    @pytest.mark.asyncio
+    async def test_generate_json_handles_none_usage(self):
+        """Token counts default to 0 when usage_metadata attributes are None."""
+        provider = GeminiProvider(api_key="test-key", model="gemini-2.5-flash")
+
+        mock_usage = MagicMock(spec=[])  # No attributes at all
+        mock_response = MagicMock()
+        mock_response.text = "{}"
+        mock_response.usage_metadata = mock_usage
+
+        mock_client = MagicMock()
+        mock_client.aio.models.generate_content = AsyncMock(
+            return_value=mock_response
+        )
+
+        mock_genai_module = MagicMock()
+        mock_genai_module.Client.return_value = mock_client
+
+        mock_types = MagicMock()
+        mock_types.Content.side_effect = lambda **kw: kw
+        mock_types.Part.side_effect = lambda **kw: kw
+        mock_types.GenerateContentConfig.side_effect = lambda **kw: kw
+
+        mock_google = MagicMock()
+        mock_google.genai = mock_genai_module
+        mock_genai_module.types = mock_types
+
+        with patch.dict(sys.modules, {
+            "google": mock_google,
+            "google.genai": mock_genai_module,
+            "google.genai.types": mock_types,
+        }):
+            text, input_tokens, output_tokens = await provider.generate_json(
+                system_prompt="test",
+                messages=[{"role": "user", "content": "test"}],
+            )
+
+        assert text == "{}"
+        assert input_tokens == 0
+        assert output_tokens == 0
--- a/docs/plans/2026-02-26-ai-autofix-gemini-design.md
+++ b/docs/plans/2026-02-26-ai-autofix-gemini-design.md
@@ -0,0 +1,209 @@
+# AI Auto-Fix & Gemini Flash Provider Design
+
+> **Date:** 2026-02-26
+> **Status:** Approved
+
+---
+
+## Overview
+
+Two combined features:
+
+1. **AI Provider Abstraction** — Add Gemini 2.5 Flash as the default AI provider with Claude as fallback, behind a unified interface.
+2. **AI Auto-Fix for Validation Errors** — When a flow fails validation, offer an AI-powered "Fix with AI" button that generates structural fixes for review.
+
+---
+
+## Section 1: AI Provider Abstraction
+
+### Design
+
+New `backend/app/core/ai_provider.py` with a unified interface:
+
+```python
+class AIProvider(ABC):
+    async def generate_json(
+        self,
+        system_prompt: str,
+        messages: list[dict],
+        max_tokens: int = 4096,
+    ) -> tuple[str, int, int]:
+        """Returns (text, input_tokens, output_tokens)"""
+```
+
+Two implementations:
+
+| Provider | Model | SDK | Role |
+|----------|-------|-----|------|
+| `GeminiProvider` | `gemini-2.5-flash` | `google-genai` | Default |
+| `AnthropicProvider` | `claude-haiku-4-5-20251001` | `anthropic` | Fallback |
+
+### Provider Selection
+
+- `get_ai_provider()` factory reads `AI_PROVIDER` env var (default: `"gemini"`)
+- Falls back to Anthropic if Gemini key is missing
+- Existing `ai_tree_generator_service.py` swaps direct Anthropic calls for `get_ai_provider()`
+
+### New Environment Variables
+
+| Variable | Default | Purpose |
+|----------|---------|---------|
+| `AI_PROVIDER` | `"gemini"` | Which provider to use (`gemini` or `anthropic`) |
+| `GOOGLE_AI_API_KEY` | — | Gemini API key |
+
+Existing `ANTHROPIC_API_KEY` remains for fallback.
+
+### Config Changes (`core/config.py`)
+
+```python
+AI_PROVIDER: str = "gemini"
+GOOGLE_AI_API_KEY: str | None = None
+AI_MODEL_GEMINI: str = "gemini-2.5-flash"
+AI_MODEL_ANTHROPIC: str = "claude-haiku-4-5-20251001"
+```
+
+---
+
+## Section 2: AI Auto-Fix Feature
+
+### Backend Endpoint
+
+**`POST /api/v1/ai/fix-tree`**
+
+Request:
+```json
+{
+  "tree_structure": { /* full tree */ },
+  "tree_name": "Router Troubleshooting",
+  "tree_type": "troubleshooting",
+  "validation_errors": [
+    {
+      "node_id": "node_abc",
+      "message": "Decision node must have at least 2 children (branches)"
+    }
+  ]
+}
+```
+
+Response:
+```json
+{
+  "fixes": [
+    {
+      "target_node_id": "node_abc",
+      "error_message": "Decision node must have at least 2 children (branches)",
+      "description": "Added second branch 'Check firmware version' with solution node",
+      "original_node": { /* snapshot before fix */ },
+      "fixed_node": { /* replacement node with corrected subtree */ }
+    }
+  ],
+  "tokens_used": { "input": 1200, "output": 800 }
+}
+```
+
+### How It Works
+
+1. For each validation error tied to a `node_id`, extract that node + its parent + siblings from the tree.
+2. Build a prompt with:
+   - The **full tree structure** serialized as a simplified outline (node titles + types + structure) for context
+   - The **specific failing node** highlighted with full JSON detail
+   - The **validation error message**
+   - Instructions: "Fix ONLY this node's structural issue. Keep all existing content. Generate domain-relevant additions that fit the flow's topic."
+3. AI returns a corrected version of that node (with children/options adjusted).
+4. Backend re-validates the fixed node before returning it.
+5. If re-validation fails, retry once with the error fed back (corrective prompt pattern).
+
+### Prompt Strategy
+
+The prompt gives the AI the full tree as a compact outline, then zooms into the failing node:
+
+```
+You are fixing a validation error in a troubleshooting flow called "Router Troubleshooting".
+
+FULL FLOW OUTLINE:
+- [decision] Is the router powered on?
+  - [action] Check power cable → [solution] Power restored
+  - [decision] Are lights blinking? ← ERROR HERE
+    - [solution] Contact ISP
+
+ERROR: Decision node "Are lights blinking?" must have at least 2 children (branches).
+
+FAILING NODE (full detail):
+{...json...}
+
+Fix this node by adding the minimum structure needed to resolve the error.
+Return ONLY the fixed node as JSON.
+```
+
+### Frontend UX
+
+1. **Trigger**: "Fix with AI" button in `ValidationSummary` — appears when there are fixable errors (structural errors with a `node_id`).
+2. **Loading state**: Button shows spinner + "Generating fixes..." — disabled during request.
+3. **Review modal** (`AIFixReviewModal`): Shows each proposed fix as a card:
+   - Error message at top
+   - Before/after view of the node change
+   - "Apply" / "Skip" buttons per fix
+   - "Apply All" button in footer
+4. **Apply**: Each accepted fix calls `updateNode(targetNodeId, fixedNode)` in the tree editor store.
+5. **Re-validate**: After applying fixes, auto-run `validate()` to confirm resolution.
+
+---
+
+## Section 3: Scope & Constraints
+
+### Fixable Errors (Auto-Fix Scope)
+
+Only structural validation errors with a `node_id`:
+- Decision node missing children/branches
+- Decision node missing options
+- Action node missing `next_node_id`
+- Dead-end decision nodes (no children)
+
+### NOT Fixable
+
+- Global checks (tree too small/large, not enough solutions) — require rethinking the whole tree
+- Content quality issues — out of scope
+- Errors without a `node_id` (root-level issues)
+
+Non-fixable errors still show in ValidationSummary but without the "Fix with AI" option.
+
+### Token Budget
+
+- Tree outline: ~50-100 tokens for a typical 15-node tree
+- Failing node detail: ~100-200 tokens
+- System prompt + instructions: ~300 tokens
+- **Total input per fix: ~500-600 tokens**
+- One API call per failing node (not batched)
+
+### Error Handling
+
+- Provider failure (rate limit, network): toast error, user can retry
+- Fix fails re-validation: "AI couldn't generate a valid fix" with retry option
+- Max 1 retry with corrective prompt per attempt
+- Both provider and fallback fail: surface error to user
+
+### Auth
+
+- Requires `engineer` role or above (`require_engineer_or_admin`)
+
+---
+
+## New Files
+
+| File | Purpose |
+|------|---------|
+| `backend/app/core/ai_provider.py` | Provider abstraction + Gemini/Anthropic implementations |
+| `backend/app/core/ai_fix_service.py` | Fix generation logic + prompt building |
+| `backend/app/api/endpoints/ai.py` | `POST /ai/fix-tree` endpoint |
+| `backend/app/schemas/ai.py` | Request/response schemas for AI endpoints |
+| `frontend/src/components/tree-editor/AIFixReviewModal.tsx` | Review modal for proposed fixes |
+
+## Modified Files
+
+| File | Change |
+|------|--------|
+| `backend/app/core/config.py` | Add Gemini config vars |
+| `backend/app/core/ai_tree_generator_service.py` | Swap Anthropic calls for provider abstraction |
+| `backend/app/api/router.py` | Register `/ai` routes |
+| `frontend/src/api/trees.ts` | Add `fixTree()` API call |
+| `frontend/src/components/tree-editor/ValidationSummary.tsx` | Add "Fix with AI" button |
--- a/docs/plans/2026-02-26-ai-autofix-gemini-plan.md
+++ b/docs/plans/2026-02-26-ai-autofix-gemini-plan.md
--- a/frontend/src/api/trees.ts
+++ b/frontend/src/api/trees.ts
@@ -1,5 +1,5 @@
 import apiClient from './client'
-import type { Tree, TreeListItem, TreeCreate, TreeUpdate, TreeFilters, TreeShareCreate, TreeShare, TreeVisibilityUpdate, TreeValidationResponse } from '@/types'
+import type { Tree, TreeListItem, TreeCreate, TreeUpdate, TreeFilters, TreeShareCreate, TreeShare, TreeVisibilityUpdate, TreeValidationResponse, AIFixTreeRequest, AIFixTreeResponse } from '@/types'

 export const treesApi = {
  async list(params?: TreeFilters): Promise<TreeListItem[]> {
@@ -65,6 +65,12 @@ export const treesApi = {
    const response = await apiClient.post<TreeValidationResponse>(`/trees/${id}/can-publish`)
    return response.data
  },
+
+  // AI auto-fix
+  async fixTree(request: AIFixTreeRequest): Promise<AIFixTreeResponse> {
+    const response = await apiClient.post<AIFixTreeResponse>('/ai/fix-tree', request)
+    return response.data
+  },
 }

 export default treesApi
--- a/frontend/src/components/tree-editor/AIFixReviewModal.tsx
+++ b/frontend/src/components/tree-editor/AIFixReviewModal.tsx
@@ -0,0 +1,170 @@
+import { useState } from 'react'
+import { X, Check, SkipForward, Sparkles, ChevronDown, ChevronUp } from 'lucide-react'
+import { cn } from '@/lib/utils'
+import type { AIFixProposal } from '@/types'
+
+interface AIFixReviewModalProps {
+  fixes: AIFixProposal[]
+  onApply: (fix: AIFixProposal) => void
+  onApplyAll: () => void
+  onClose: () => void
+}
+
+export function AIFixReviewModal({ fixes, onApply, onApplyAll, onClose }: AIFixReviewModalProps) {
+  const [appliedIds, setAppliedIds] = useState<Set<string>>(new Set())
+  const [skippedIds, setSkippedIds] = useState<Set<string>>(new Set())
+  const [expandedIds, setExpandedIds] = useState<Set<string>>(new Set(fixes.map(f => f.target_node_id)))
+
+  const handleApply = (fix: AIFixProposal) => {
+    onApply(fix)
+    setAppliedIds(prev => new Set(prev).add(fix.target_node_id))
+  }
+
+  const handleSkip = (fix: AIFixProposal) => {
+    setSkippedIds(prev => new Set(prev).add(fix.target_node_id))
+  }
+
+  const toggleExpanded = (id: string) => {
+    setExpandedIds(prev => {
+      const next = new Set(prev)
+      if (next.has(id)) next.delete(id)
+      else next.add(id)
+      return next
+    })
+  }
+
+  const pendingFixes = fixes.filter(
+    f => !appliedIds.has(f.target_node_id) && !skippedIds.has(f.target_node_id)
+  )
+  const allHandled = pendingFixes.length === 0
+
+  return (
+    <div className="fixed inset-0 z-50 flex items-center justify-center bg-black/80 backdrop-blur-sm p-4">
+      <div className="relative flex h-[80vh] w-full max-w-2xl flex-col bg-card border border-border rounded-2xl shadow-lg">
+        {/* Header */}
+        <div className="flex items-center justify-between border-b border-border px-6 py-4">
+          <div className="flex items-center gap-2">
+            <Sparkles className="h-5 w-5 text-primary" />
+            <h2 className="text-lg font-semibold text-foreground">
+              AI Fix Proposals ({fixes.length})
+            </h2>
+          </div>
+          <button
+            onClick={onClose}
+            className="rounded-md p-1 text-muted-foreground hover:bg-accent hover:text-foreground"
+          >
+            <X className="h-5 w-5" />
+          </button>
+        </div>
+
+        {/* Body */}
+        <div className="flex-1 overflow-y-auto p-4 space-y-3">
+          {fixes.map((fix) => {
+            const isApplied = appliedIds.has(fix.target_node_id)
+            const isSkipped = skippedIds.has(fix.target_node_id)
+            const isExpanded = expandedIds.has(fix.target_node_id)
+
+            return (
+              <div
+                key={fix.target_node_id}
+                className={cn(
+                  'rounded-lg border p-4',
+                  isApplied
+                    ? 'border-emerald-400/30 bg-emerald-400/5'
+                    : isSkipped
+                      ? 'border-border bg-accent/30 opacity-60'
+                      : 'border-border bg-card'
+                )}
+              >
+                {/* Fix header */}
+                <div className="flex items-start justify-between gap-3">
+                  <div className="flex-1">
+                    <p className="text-sm text-red-400 mb-1">{fix.error_message}</p>
+                    <p className="text-sm text-foreground">{fix.description}</p>
+                    <p className="text-xs text-muted-foreground mt-1">
+                      Node: {fix.target_node_id}
+                    </p>
+                  </div>
+                  {isApplied && (
+                    <span className="flex items-center gap-1 rounded-full bg-emerald-400/10 px-2 py-1 text-xs text-emerald-400">
+                      <Check className="h-3 w-3" /> Applied
+                    </span>
+                  )}
+                  {isSkipped && (
+                    <span className="text-xs text-muted-foreground">Skipped</span>
+                  )}
+                </div>
+
+                {/* Expand/collapse detail */}
+                {!isApplied && !isSkipped && (
+                  <>
+                    <button
+                      onClick={() => toggleExpanded(fix.target_node_id)}
+                      className="mt-2 flex items-center gap-1 text-xs text-muted-foreground hover:text-foreground"
+                    >
+                      {isExpanded ? <ChevronUp className="h-3 w-3" /> : <ChevronDown className="h-3 w-3" />}
+                      {isExpanded ? 'Hide' : 'Show'} details
+                    </button>
+
+                    {isExpanded && (
+                      <div className="mt-3 grid grid-cols-2 gap-3">
+                        <div>
+                          <p className="text-xs font-medium text-muted-foreground mb-1">Before</p>
+                          <pre className="overflow-x-auto rounded bg-accent/50 p-2 text-xs text-muted-foreground max-h-48 overflow-y-auto">
+                            {JSON.stringify(fix.original_node, null, 2)}
+                          </pre>
+                        </div>
+                        <div>
+                          <p className="text-xs font-medium text-emerald-400 mb-1">After</p>
+                          <pre className="overflow-x-auto rounded bg-emerald-400/5 p-2 text-xs text-foreground max-h-48 overflow-y-auto">
+                            {JSON.stringify(fix.fixed_node, null, 2)}
+                          </pre>
+                        </div>
+                      </div>
+                    )}
+
+                    {/* Action buttons */}
+                    <div className="mt-3 flex gap-2">
+                      <button
+                        onClick={() => handleApply(fix)}
+                        className="flex items-center gap-1 rounded-md bg-gradient-brand px-3 py-1.5 text-xs font-medium text-white shadow-sm shadow-primary/20 hover:opacity-90"
+                      >
+                        <Check className="h-3 w-3" />
+                        Apply
+                      </button>
+                      <button
+                        onClick={() => handleSkip(fix)}
+                        className="flex items-center gap-1 rounded-md border border-border px-3 py-1.5 text-xs font-medium text-muted-foreground hover:bg-accent hover:text-foreground"
+                      >
+                        <SkipForward className="h-3 w-3" />
+                        Skip
+                      </button>
+                    </div>
+                  </>
+                )}
+              </div>
+            )
+          })}
+        </div>
+
+        {/* Footer */}
+        <div className="flex items-center justify-between border-t border-border px-6 py-4">
+          <button
+            onClick={onClose}
+            className="rounded-md border border-border px-4 py-2 text-sm font-medium text-muted-foreground hover:bg-accent hover:text-foreground"
+          >
+            {allHandled ? 'Done' : 'Cancel'}
+          </button>
+          {!allHandled && (
+            <button
+              onClick={onApplyAll}
+              className="rounded-md bg-gradient-brand px-4 py-2 text-sm font-medium text-white shadow-lg shadow-primary/20 hover:opacity-90"
+            >
+              Apply All ({pendingFixes.length})
+            </button>
+          )}
+        </div>
+      </div>
+    </div>
+  )
+}
--- a/frontend/src/components/tree-editor/ValidationSummary.tsx
+++ b/frontend/src/components/tree-editor/ValidationSummary.tsx
@@ -1,14 +1,16 @@
 import { useState } from 'react'
-import { AlertCircle, AlertTriangle, ChevronDown, ChevronUp } from 'lucide-react'
+import { AlertCircle, AlertTriangle, ChevronDown, ChevronUp, Sparkles, Loader2 } from 'lucide-react'
 import { cn } from '@/lib/utils'
 import type { ValidationError } from '@/store/treeEditorStore'

 interface ValidationSummaryProps {
  errors: ValidationError[]
  onSelectNode: (nodeId: string) => void
+  onFixWithAI?: () => void
+  isFixing?: boolean
 }

-export function ValidationSummary({ errors, onSelectNode }: ValidationSummaryProps) {
+export function ValidationSummary({ errors, onSelectNode, onFixWithAI, isFixing }: ValidationSummaryProps) {
  const [isExpanded, setIsExpanded] = useState(true)

  const errorItems = errors.filter(e => e.severity === 'error')
@@ -22,6 +24,8 @@ export function ValidationSummary({ errors, onSelectNode }: ValidationSummaryPro
    }
  }

+  const hasFixableErrors = errorItems.some(e => e.nodeId)
+
  return (
    <div
      className={cn(
@@ -32,14 +36,16 @@ export function ValidationSummary({ errors, onSelectNode }: ValidationSummaryPro
      )}
    >
      {/* Header */}
-      <button
-        onClick={() => setIsExpanded(!isExpanded)}
+      <div
        className={cn(
-          'flex w-full items-center justify-between p-3 text-left transition-colors hover:bg-accent',
+          'flex w-full items-center justify-between p-3 transition-colors',
          errorItems.length > 0 ? 'text-red-400' : 'text-yellow-400'
        )}
      >
-        <div className="flex items-center gap-2">
+        <button
+          onClick={() => setIsExpanded(!isExpanded)}
+          className="flex items-center gap-2 text-left hover:opacity-80"
+        >
          {errorItems.length > 0 ? (
            <AlertCircle className="h-5 w-5" />
          ) : (
@@ -58,9 +64,35 @@ export function ValidationSummary({ errors, onSelectNode }: ValidationSummaryPro
              </>
            )}
          </span>
-        </div>
-        {isExpanded ? <ChevronUp className="h-4 w-4" /> : <ChevronDown className="h-4 w-4" />}
-      </button>
+          {isExpanded ? <ChevronUp className="h-4 w-4" /> : <ChevronDown className="h-4 w-4" />}
+        </button>
+
+        {/* Fix with AI button */}
+        {onFixWithAI && hasFixableErrors && (
+          <button
+            onClick={onFixWithAI}
+            disabled={isFixing}
+            className={cn(
+              'flex items-center gap-1.5 rounded-md px-3 py-1 text-xs font-medium transition-colors',
+              isFixing
+                ? 'bg-primary/10 text-primary cursor-wait'
+                : 'bg-gradient-brand text-white shadow-sm shadow-primary/20 hover:opacity-90'
+            )}
+          >
+            {isFixing ? (
+              <>
+                <Loader2 className="h-3 w-3 animate-spin" />
+                Generating fixes...
+              </>
+            ) : (
+              <>
+                <Sparkles className="h-3 w-3" />
+                Fix with AI
+              </>
+            )}
+          </button>
+        )}
+      </div>

      {/* Error/Warning List */}
      {isExpanded && (
--- a/frontend/src/pages/TreeEditorPage.tsx
+++ b/frontend/src/pages/TreeEditorPage.tsx
@@ -5,10 +5,11 @@ import { Undo2, Redo2, Save, CheckCircle2, Monitor, FileText, Code2, LayoutList,
 import { getMonacoEditor } from '@/components/tree-editor/code-mode'
 import { treesApi } from '@/api/trees'
 import { treeMarkdownApi } from '@/api/treeMarkdown'
-import type { TreeCreate, TreeUpdate, TreeStatus, TreeStructure } from '@/types'
+import type { TreeCreate, TreeUpdate, TreeStatus, TreeStructure, AIFixProposal } from '@/types'
 import { useTreeEditorStore, useTreeEditorTemporal } from '@/store/treeEditorStore'
 import { TreeEditorLayout } from '@/components/tree-editor/TreeEditorLayout'
 import { ValidationSummary } from '@/components/tree-editor/ValidationSummary'
+import { AIFixReviewModal } from '@/components/tree-editor/AIFixReviewModal'
 import { useKeyboardShortcuts } from '@/hooks/useKeyboardShortcuts'
 import { usePermissions } from '@/hooks/usePermissions'
 import { Spinner } from '@/components/common/Spinner'
@@ -58,6 +59,8 @@ export function TreeEditorPage() {
  const [showAnalytics, setShowAnalytics] = useState(false)
  const [isMetadataOpen, setIsMetadataOpen] = useState(false)
  const [editingNodeId, setEditingNodeId] = useState<string | null>(null)
+  const [isFixing, setIsFixing] = useState(false)
+  const [fixProposals, setFixProposals] = useState<AIFixProposal[] | null>(null)

  // Mobile detection
  const [isMobile, setIsMobile] = useState(false)
@@ -217,6 +220,54 @@ export function TreeEditorPage() {
    selectNode(nodeId)
  }

+  const handleFixWithAI = async () => {
+    const store = useTreeEditorStore.getState()
+    if (!store.treeStructure) return
+
+    const fixableErrors = store.validationErrors
+      .filter(e => e.severity === 'error' && e.nodeId)
+      .map(e => ({ node_id: e.nodeId!, message: e.message }))
+
+    if (fixableErrors.length === 0) return
+
+    setIsFixing(true)
+    try {
+      const result = await treesApi.fixTree({
+        tree_structure: store.treeStructure as unknown as Record<string, unknown>,
+        tree_name: store.name,
+        tree_type: 'troubleshooting',
+        validation_errors: fixableErrors,
+      })
+      if (result.fixes.length > 0) {
+        setFixProposals(result.fixes)
+      } else {
+        toast.info('AI could not generate fixes for these errors')
+      }
+    } catch {
+      toast.error('Failed to generate AI fixes. Please try again.')
+    } finally {
+      setIsFixing(false)
+    }
+  }
+
+  const handleApplyFix = (fix: AIFixProposal) => {
+    updateNode(fix.target_node_id, fix.fixed_node as Partial<TreeStructure>)
+  }
+
+  const handleApplyAllFixes = () => {
+    if (!fixProposals) return
+    for (const fix of fixProposals) {
+      handleApplyFix(fix)
+    }
+    setFixProposals(null)
+    setTimeout(() => { validate() }, 100)
+  }
+
+  const handleCloseFixModal = () => {
+    setFixProposals(null)
+    validate()
+  }
+
  const handleNodeSelect = useCallback((nodeId: string | null) => {
    if (nodeId) {
      setIsMetadataOpen(false) // close metadata when opening node editor
@@ -685,6 +736,8 @@ export function TreeEditorPage() {
          <ValidationSummary
            errors={validationErrors}
            onSelectNode={handleSelectNode}
+            onFixWithAI={handleFixWithAI}
+            isFixing={isFixing}
          />
        </div>
      )}
@@ -705,6 +758,16 @@ export function TreeEditorPage() {
          <FlowAnalyticsPanel treeId={id} />
        </div>
      )}
+
+      {/* AI Fix Review Modal */}
+      {fixProposals && (
+        <AIFixReviewModal
+          fixes={fixProposals}
+          onApply={handleApplyFix}
+          onApplyAll={handleApplyAllFixes}
+          onClose={handleCloseFixModal}
+        />
+      )}
    </div>
  )
 }
--- a/frontend/src/types/ai-fix.ts
+++ b/frontend/src/types/ai-fix.ts
@@ -0,0 +1,24 @@
+export interface AIFixValidationError {
+  node_id: string
+  message: string
+}
+
+export interface AIFixProposal {
+  target_node_id: string
+  error_message: string
+  description: string
+  original_node: Record<string, unknown>
+  fixed_node: Record<string, unknown>
+}
+
+export interface AIFixTreeRequest {
+  tree_structure: Record<string, unknown>
+  tree_name: string
+  tree_type: 'troubleshooting' | 'procedural' | 'maintenance'
+  validation_errors: AIFixValidationError[]
+}
+
+export interface AIFixTreeResponse {
+  fixes: AIFixProposal[]
+  tokens_used: { input: number; output: number }
+}
--- a/frontend/src/types/index.ts
+++ b/frontend/src/types/index.ts
@@ -45,3 +45,10 @@ export type {
  AIAssembleResponse,
  AIWizardPhase,
 } from './ai'
+
+export type {
+  AIFixTreeRequest,
+  AIFixTreeResponse,
+  AIFixProposal,
+  AIFixValidationError,
+} from './ai-fix'