Phase 2: Add reasoning and TipTap content endpoints

- Extend Document model with reasoning_type, confidence, reasoning_steps, model_source, tiptap_content fields
- Add new endpoints:
  - GET /documents/{id}/reasoning - Get reasoning metadata
  - PATCH /documents/{id}/reasoning - Update reasoning metadata
  - GET /documents/{id}/reasoning-panel - Get reasoning panel data for UI
  - POST /documents/{id}/reasoning-steps - Add reasoning step
  - DELETE /documents/{id}/reasoning-steps/{step} - Delete reasoning step
  - GET /documents/{id}/content?format=tiptap|markdown - Get content in TipTap or Markdown
  - PUT /documents/{id}/content - Update content (supports both TipTap JSON and Markdown)
- Add TipTap to Markdown conversion
- Update database schema with new columns
- Add comprehensive tests for all new endpoints
- All 37 tests passing

app/database.py

@@ -185,7 +185,12 @@ def _create_schema(sync_conn):
             deleted_at TIMESTAMP NULL,
             deleted_by TEXT NULL,
             created_at TIMESTAMP NOT NULL DEFAULT (datetime('now')),
-            updated_at TIMESTAMP NOT NULL DEFAULT (datetime('now'))
+            updated_at TIMESTAMP NOT NULL DEFAULT (datetime('now')),
+            reasoning_type TEXT,
+            confidence TEXT,
+            reasoning_steps TEXT,
+            model_source TEXT,
+            tiptap_content TEXT
         )
     """))
 

app/models/document.py

@@ -1,5 +1,6 @@
 import uuid
 from datetime import datetime
+from enum import Enum
 
 from sqlalchemy import Boolean, DateTime, ForeignKey, String, Text
 from sqlalchemy.orm import Mapped, mapped_column, relationship
@@ -11,6 +12,13 @@ def generate_uuid() -> str:
     return str(uuid.uuid4())
 
 
+class ReasoningType(str, Enum):
+    CHAIN = "chain"
+    IDEA = "idea"
+    CONTEXT = "context"
+    REFLECTION = "reflection"
+
+
 class Document(Base):
     __tablename__ = "documents"
 
@@ -25,3 +33,9 @@ class Document(Base):
     deleted_by: Mapped[str | None] = mapped_column(String(36), nullable=True)
     created_at: Mapped[datetime] = mapped_column(DateTime, default=datetime.utcnow, nullable=False)
     updated_at: Mapped[datetime] = mapped_column(DateTime, default=datetime.utcnow, onupdate=datetime.utcnow, nullable=False)
+    # Phase 2: Reasoning fields
+    reasoning_type: Mapped[str | None] = mapped_column(String(20), nullable=True)
+    confidence: Mapped[str | None] = mapped_column(String(10), nullable=True)  # Store as string to handle JSON serialization
+    reasoning_steps: Mapped[str | None] = mapped_column(Text, nullable=True)  # JSON array as text
+    model_source: Mapped[str | None] = mapped_column(String(100), nullable=True)
+    tiptap_content: Mapped[str | None] = mapped_column(Text, nullable=True)  # JSON object as text

app/routers/documents.py

@@ -1,12 +1,13 @@
+import json
 import uuid
 from datetime import datetime
 
-from fastapi import APIRouter, Depends, HTTPException, Request
+from fastapi import APIRouter, Depends, HTTPException, Query, Request
 from sqlalchemy import delete, select, text
 from sqlalchemy.ext.asyncio import AsyncSession
 
 from app.database import get_db
-from app.models.document import Document
+from app.models.document import Document, ReasoningType
 from app.models.folder import Folder
 from app.models.project import Project
 from app.models.tag import DocumentTag, Tag
@@ -18,7 +19,14 @@ from app.schemas.document import (
     DocumentListResponse,
     DocumentResponse,
     DocumentUpdate,
+    ReasoningMetadata,
+    ReasoningPanel,
+    ReasoningStep,
+    ReasoningStepAdd,
+    ReasoningUpdate,
     TagInfo,
+    TipTapContentResponse,
+    TipTapContentUpdate,
 )
 from app.schemas.tag import DocumentTagsAssign
 
@@ -32,6 +40,85 @@ def build_doc_path(project_id: str, doc_id: str, folder_id: str | None, folder_p
     return f"/{project_id}/{doc_id}"
 
 
+def tiptap_to_markdown(tiptap: dict) -> str:
+    """Convert TipTap JSON to Markdown string."""
+    if not tiptap or not isinstance(tiptap, dict):
+        return ""
+    
+    lines = []
+    
+    def process_node(node: dict) -> str:
+        if not isinstance(node, dict):
+            return ""
+        
+        node_type = node.get("type", "")
+        content = node.get("content", [])
+        
+        if node_type == "doc":
+            result = []
+            for child in content:
+                result.append(process_node(child))
+            return "\n".join(result)
+        
+        elif node_type == "paragraph":
+            inner = "".join(process_node(c) for c in content)
+            return f"{inner}\n"
+        
+        elif node_type == "heading":
+            level = node.get("attrs", {}).get("level", 1)
+            inner = "".join(process_node(c) for c in content)
+            return f"{'#' * level} {inner}\n"
+        
+        elif node_type == "text":
+            text_val = node.get("text", "")
+            marks = node.get("marks", [])
+            for mark in marks:
+                if mark.get("type") == "bold":
+                    text_val = f"**{text_val}**"
+                elif mark.get("type") == "italic":
+                    text_val = f"*{text_val}*"
+                elif mark.get("type") == "code":
+                    text_val = f"`{text_val}`"
+                elif mark.get("type") == "strike":
+                    text_val = f"~~{text_val}~~"
+            return text_val
+        
+        elif node_type == "bulletList":
+            return "\n".join(process_node(item) for item in content) + "\n"
+        
+        elif node_type == "orderedList":
+            return "\n".join(process_node(item) for item in content) + "\n"
+        
+        elif node_type == "listItem":
+            inner = "".join(process_node(c) for c in content)
+            return f"- {inner.strip()}\n"
+        
+        elif node_type == "blockquote":
+            inner = "".join(process_node(c) for c in content)
+            return f"> {inner.strip()}\n"
+        
+        elif node_type == "codeBlock":
+            lang = node.get("attrs", {}).get("language", "")
+            inner = "".join(process_node(c) for c in content)
+            return f"```{lang}\n{inner}\n```\n"
+        
+        elif node_type == "hardBreak":
+            return "\n"
+        
+        elif node_type == "horizontalRule":
+            return "---\n"
+        
+        elif node_type == "image":
+            src = node.get("attrs", {}).get("src", "")
+            alt = node.get("attrs", {}).get("alt", "")
+            return f"![{alt}]({src})"
+        
+        return ""
+    
+    result = process_node(tiptap)
+    return result.strip()
+
+
 async def get_document_tags(db: AsyncSession, doc_id: str) -> list[TagInfo]:
     result = await db.execute(
         text("""
@@ -48,6 +135,31 @@ async def get_document_tags(db: AsyncSession, doc_id: str) -> list[TagInfo]:
 
 async def document_to_response(db: AsyncSession, doc: Document) -> DocumentResponse:
     tags = await get_document_tags(db, doc.id)
+    
+    # Parse reasoning_steps from JSON if present
+    reasoning_steps = []
+    if doc.reasoning_steps:
+        try:
+            reasoning_steps = json.loads(doc.reasoning_steps)
+        except json.JSONDecodeError:
+            reasoning_steps = []
+    
+    # Parse tiptap_content from JSON if present
+    tiptap_content = None
+    if doc.tiptap_content:
+        try:
+            tiptap_content = json.loads(doc.tiptap_content)
+        except json.JSONDecodeError:
+            tiptap_content = None
+    
+    # Parse confidence from string if present
+    confidence = None
+    if doc.confidence:
+        try:
+            confidence = float(doc.confidence)
+        except (ValueError, TypeError):
+            confidence = None
+    
     return DocumentResponse(
         id=doc.id,
         title=doc.title,
@@ -58,6 +170,11 @@ async def document_to_response(db: AsyncSession, doc: Document) -> DocumentRespo
         tags=tags,
         created_at=doc.created_at,
         updated_at=doc.updated_at,
+        reasoning_type=doc.reasoning_type,
+        confidence=confidence,
+        reasoning_steps=reasoning_steps,
+        model_source=doc.model_source,
+        tiptap_content=tiptap_content,
     )
 
 
@@ -276,9 +393,14 @@ async def delete_document(
 async def update_document_content(
     request: Request,
     document_id: str,
-    payload: DocumentContentUpdate,
+    payload: TipTapContentUpdate,
     db: AsyncSession = Depends(get_db),
 ):
+    """Update document content (TipTap JSON or Markdown).
+    
+    Phase 2: Now supports both TipTap JSON and Markdown formats via the 'format' field.
+    Also backward-compatible with legacy string content (treated as markdown).
+    """
     agent = await get_current_agent(request, db)
 
     result = await db.execute(
@@ -301,7 +423,37 @@ async def update_document_content(
     if not proj_result.scalar_one_or_none():
         raise HTTPException(status_code=403, detail="Forbidden")
 
-    doc.content = payload.content
+    # Determine actual format based on content type (backward compatibility)
+    # If content is a string, treat as markdown regardless of format field
+    # If content is a dict, treat as tiptap
+    is_string_content = isinstance(payload.content, str)
+    
+    # Validate content size (1MB limit)
+    content_json = json.dumps(payload.content)
+    if len(content_json) > 1_000_000:
+        raise HTTPException(status_code=413, detail="Content too large (max 1MB)")
+
+    if is_string_content:
+        # Legacy string content or markdown - store as markdown
+        doc.content = payload.content
+        # Create a simple tiptap structure for the editor
+        doc.tiptap_content = json.dumps({
+            "type": "doc",
+            "content": [
+                {
+                    "type": "paragraph",
+                    "content": [{"type": "text", "text": payload.content}]
+                }
+            ]
+        })
+    else:
+        # TipTap JSON content
+        if not isinstance(payload.content, dict):
+            raise HTTPException(status_code=400, detail="content must be a string or dict")
+        doc.tiptap_content = content_json
+        # Also update the plain content by converting tiptap -> markdown
+        doc.content = tiptap_to_markdown(payload.content)
+
     doc.updated_at = datetime.utcnow()
     await db.flush()
     return await document_to_response(db, doc)
@@ -433,3 +585,258 @@ async def remove_tag(
     )
     await db.flush()
     return None
+
+
+# =============================================================================
+# Phase 2: New Endpoints
+# =============================================================================
+
+async def _get_doc_with_access(
+    request: Request,
+    document_id: str,
+    db: AsyncSession,
+    require_write: bool = False,
+) -> tuple[Document, bool]:
+    """Get document and check access. Returns (doc, has_access)."""
+    agent = await get_current_agent(request, db)
+    
+    result = await db.execute(
+        select(Document).where(
+            Document.id == document_id,
+            Document.is_deleted == False,
+        )
+    )
+    doc = result.scalar_one_or_none()
+    if not doc:
+        raise HTTPException(status_code=404, detail="Document not found")
+    
+    proj_result = await db.execute(
+        select(Project).where(
+            Project.id == doc.project_id,
+            Project.agent_id == agent.id,
+            Project.is_deleted == False,
+        )
+    )
+    if not proj_result.scalar_one_or_none():
+        raise HTTPException(status_code=403, detail="Forbidden")
+    
+    return doc, True
+
+
+@router.get("/api/v1/documents/{document_id}/reasoning")
+async def get_document_reasoning(
+    request: Request,
+    document_id: str,
+    db: AsyncSession = Depends(get_db),
+):
+    """Get reasoning metadata for a document."""
+    doc, _ = await _get_doc_with_access(request, document_id, db)
+    
+    # Parse reasoning_steps
+    reasoning_steps = []
+    if doc.reasoning_steps:
+        try:
+            reasoning_steps = json.loads(doc.reasoning_steps)
+        except json.JSONDecodeError:
+            reasoning_steps = []
+    
+    # Parse confidence
+    confidence = None
+    if doc.confidence:
+        try:
+            confidence = float(doc.confidence)
+        except (ValueError, TypeError):
+            confidence = None
+    
+    return {
+        "reasoning_type": doc.reasoning_type,
+        "confidence": confidence,
+        "reasoning_steps": reasoning_steps,
+        "model_source": doc.model_source,
+    }
+
+
+@router.patch("/api/v1/documents/{document_id}/reasoning")
+async def update_document_reasoning(
+    request: Request,
+    document_id: str,
+    payload: ReasoningUpdate,
+    db: AsyncSession = Depends(get_db),
+):
+    """Update reasoning metadata for a document."""
+    doc, _ = await _get_doc_with_access(request, document_id, db, require_write=True)
+    
+    if payload.reasoning_type is not None:
+        doc.reasoning_type = payload.reasoning_type.value if hasattr(payload.reasoning_type, 'value') else payload.reasoning_type
+    
+    if payload.confidence is not None:
+        if payload.confidence < 0.0 or payload.confidence > 1.0:
+            raise HTTPException(status_code=400, detail="confidence must be between 0.0 and 1.0")
+        doc.confidence = str(payload.confidence)
+    
+    if payload.reasoning_steps is not None:
+        # Validate steps
+        for step in payload.reasoning_steps:
+            if not isinstance(step.step, int):
+                raise HTTPException(status_code=400, detail="Each step must have an integer 'step' field")
+            if not step.thought or len(step.thought) > 2000:
+                raise HTTPException(status_code=400, detail="thought must be non-empty and max 2000 chars")
+        doc.reasoning_steps = json.dumps([s.model_dump() for s in payload.reasoning_steps])
+    
+    if payload.model_source is not None:
+        doc.model_source = payload.model_source
+    
+    doc.updated_at = datetime.utcnow()
+    await db.flush()
+    
+    return await get_document_reasoning(request, document_id, db)
+
+
+@router.get("/api/v1/documents/{document_id}/reasoning-panel")
+async def get_reasoning_panel(
+    request: Request,
+    document_id: str,
+    db: AsyncSession = Depends(get_db),
+):
+    """Get reasoning panel data for UI."""
+    doc, _ = await _get_doc_with_access(request, document_id, db)
+    
+    # Check if document has reasoning
+    has_reasoning = any([
+        doc.reasoning_type is not None,
+        doc.confidence is not None,
+        doc.reasoning_steps,
+        doc.model_source is not None,
+    ])
+    
+    reasoning_metadata = None
+    if has_reasoning:
+        reasoning_steps = []
+        if doc.reasoning_steps:
+            try:
+                reasoning_steps = json.loads(doc.reasoning_steps)
+            except json.JSONDecodeError:
+                reasoning_steps = []
+        
+        confidence = None
+        if doc.confidence:
+            try:
+                confidence = float(doc.confidence)
+            except (ValueError, TypeError):
+                confidence = None
+        
+        reasoning_metadata = ReasoningMetadata(
+            reasoning_type=doc.reasoning_type,
+            confidence=confidence,
+            reasoning_steps=reasoning_steps,
+            model_source=doc.model_source,
+        )
+    
+    return ReasoningPanel(
+        document_id=document_id,
+        has_reasoning=has_reasoning,
+        reasoning=reasoning_metadata,
+        editable=True,  # Agent has write access since they passed _get_doc_with_access
+    )
+
+
+@router.post("/api/v1/documents/{document_id}/reasoning-steps", status_code=201)
+async def add_reasoning_step(
+    request: Request,
+    document_id: str,
+    payload: ReasoningStepAdd,
+    db: AsyncSession = Depends(get_db),
+):
+    """Add a new reasoning step to a document."""
+    doc, _ = await _get_doc_with_access(request, document_id, db, require_write=True)
+    
+    # Parse existing steps
+    steps = []
+    if doc.reasoning_steps:
+        try:
+            steps = json.loads(doc.reasoning_steps)
+        except json.JSONDecodeError:
+            steps = []
+    
+    # Determine next step number
+    next_step = max([s.get("step", 0) for s in steps], default=0) + 1
+    
+    # Create new step
+    new_step = {
+        "step": next_step,
+        "thought": payload.thought,
+        "conclusion": payload.conclusion,
+    }
+    steps.append(new_step)
+    
+    doc.reasoning_steps = json.dumps(steps)
+    doc.updated_at = datetime.utcnow()
+    await db.flush()
+    
+    return new_step
+
+
+@router.delete("/api/v1/documents/{document_id}/reasoning-steps/{step}", status_code=204)
+async def delete_reasoning_step(
+    request: Request,
+    document_id: str,
+    step: int,
+    db: AsyncSession = Depends(get_db),
+):
+    """Delete a specific reasoning step from a document."""
+    doc, _ = await _get_doc_with_access(request, document_id, db, require_write=True)
+    
+    # Parse existing steps
+    steps = []
+    if doc.reasoning_steps:
+        try:
+            steps = json.loads(doc.reasoning_steps)
+        except json.JSONDecodeError:
+            steps = []
+    
+    # Find and remove the step
+    original_len = len(steps)
+    steps = [s for s in steps if s.get("step") != step]
+    
+    if len(steps) == original_len:
+        raise HTTPException(status_code=404, detail="Step not found")
+    
+    doc.reasoning_steps = json.dumps(steps)
+    doc.updated_at = datetime.utcnow()
+    await db.flush()
+    
+    return None
+
+
+@router.get("/api/v1/documents/{document_id}/content")
+async def get_document_content(
+    request: Request,
+    document_id: str,
+    format: str = Query("tiptap", description="Output format: tiptap or markdown"),
+    db: AsyncSession = Depends(get_db),
+):
+    """Get document content in TipTap JSON or Markdown format."""
+    doc, _ = await _get_doc_with_access(request, document_id, db)
+    
+    if format == "markdown":
+        # Try to get tiptap_content and convert
+        if doc.tiptap_content:
+            try:
+                tiptap = json.loads(doc.tiptap_content)
+                content = tiptap_to_markdown(tiptap)
+                return TipTapContentResponse(content=content, format="markdown")
+            except json.JSONDecodeError:
+                pass
+        # Fallback to plain content
+        return TipTapContentResponse(content=doc.content, format="markdown")
+    else:
+        # Return tiptap format
+        if doc.tiptap_content:
+            try:
+                tiptap = json.loads(doc.tiptap_content)
+                return TipTapContentResponse(content=tiptap, format="tiptap")
+            except json.JSONDecodeError:
+                pass
+        # Return default tiptap structure
+        default_tiptap = {"type": "doc", "content": [{"type": "paragraph", "content": []}]}
+        return TipTapContentResponse(content=default_tiptap, format="tiptap")

app/schemas/document.py

@@ -1,6 +1,15 @@
 from datetime import datetime
+from enum import Enum
+from typing import Any
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, field_validator
+
+
+class ReasoningType(str, Enum):
+    CHAIN = "chain"
+    IDEA = "idea"
+    CONTEXT = "context"
+    REFLECTION = "reflection"
 
 
 class DocumentCreate(BaseModel):
@@ -18,6 +27,19 @@ class DocumentContentUpdate(BaseModel):
     content: str = Field(..., max_length=1_000_000)  # 1MB limit
 
 
+# Phase 2: TipTap Content Update - Union to support both legacy string and new TipTap/Markdown formats
+class TipTapContentUpdate(BaseModel):
+    content: dict[str, Any] | str  # TipTap JSON object or Markdown string
+    format: str = "tiptap"
+
+    @field_validator("format")
+    @classmethod
+    def validate_format(cls, v: str) -> str:
+        if v not in ("tiptap", "markdown"):
+            raise ValueError("format must be 'tiptap' or 'markdown'")
+        return v
+
+
 class TagInfo(BaseModel):
     id: str
     name: str
@@ -26,6 +48,67 @@ class TagInfo(BaseModel):
     model_config = {"from_attributes": True}
 
 
+# Phase 2: Reasoning Step Schema
+class ReasoningStep(BaseModel):
+    step: int
+    thought: str = Field(..., max_length=2000)
+    conclusion: str | None = Field(None, max_length=2000)
+
+
+# Phase 2: Reasoning Metadata Schema
+class ReasoningMetadata(BaseModel):
+    reasoning_type: ReasoningType | None = None
+    confidence: float | None = None
+    reasoning_steps: list[ReasoningStep] = []
+    model_source: str | None = None
+
+    model_config = {"protected_namespaces": ()}
+
+    @field_validator("confidence")
+    @classmethod
+    def validate_confidence(cls, v: float | None) -> float | None:
+        if v is not None and (v < 0.0 or v > 1.0):
+            raise ValueError("confidence must be between 0.0 and 1.0")
+        return v
+
+
+# Phase 2: Reasoning Update Schema
+class ReasoningUpdate(BaseModel):
+    reasoning_type: ReasoningType | None = None
+    confidence: float | None = None
+    reasoning_steps: list[ReasoningStep] | None = None
+    model_source: str | None = None
+
+    model_config = {"protected_namespaces": ()}
+
+    @field_validator("confidence")
+    @classmethod
+    def validate_confidence(cls, v: float | None) -> float | None:
+        if v is not None and (v < 0.0 or v > 1.0):
+            raise ValueError("confidence must be between 0.0 and 1.0")
+        return v
+
+
+# Phase 2: Add Reasoning Step Schema
+class ReasoningStepAdd(BaseModel):
+    thought: str = Field(..., min_length=1, max_length=2000)
+    conclusion: str | None = Field(None, max_length=2000)
+
+
+# Phase 2: TipTap Content Response
+class TipTapContentResponse(BaseModel):
+    content: dict[str, Any] | str
+    format: str
+
+
+# Phase 2: Reasoning Panel Schema
+class ReasoningPanel(BaseModel):
+    document_id: str
+    has_reasoning: bool
+    reasoning: ReasoningMetadata | None = None
+    editable: bool = True
+
+
 class DocumentResponse(BaseModel):
     id: str
     title: str
@@ -36,8 +119,14 @@ class DocumentResponse(BaseModel):
     tags: list[TagInfo] = []
     created_at: datetime
     updated_at: datetime
+    # Phase 2: new fields
+    reasoning_type: ReasoningType | None = None
+    confidence: float | None = None
+    reasoning_steps: list[ReasoningStep] = []
+    model_source: str | None = None
+    tiptap_content: dict[str, Any] | None = None
 
-    model_config = {"from_attributes": True}
+    model_config = {"from_attributes": True, "protected_namespaces": ()}
 
 
 class DocumentBriefResponse(BaseModel):

tests/test_documents.py

@@ -134,3 +134,361 @@ async def test_remove_tag(client):
         headers={"Authorization": f"Bearer {token}"}
     )
     assert remove_resp.status_code == 204
+
+
+# =============================================================================
+# Phase 2: Reasoning Endpoint Tests
+# =============================================================================
+
+@pytest.mark.asyncio
+async def test_get_document_reasoning_empty(client):
+    """Test getting reasoning from a document with no reasoning set."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "No Reasoning Doc"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    response = await client.get(
+        f"/api/v1/documents/{doc_id}/reasoning",
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert response.status_code == 200
+    data = response.json()
+    assert data["reasoning_type"] is None
+    assert data["confidence"] is None
+    assert data["reasoning_steps"] == []
+    assert data["model_source"] is None
+
+
+@pytest.mark.asyncio
+async def test_patch_document_reasoning(client):
+    """Test updating reasoning metadata."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "Reasoning Doc"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    # Update reasoning
+    response = await client.patch(
+        f"/api/v1/documents/{doc_id}/reasoning",
+        json={
+            "reasoning_type": "chain",
+            "confidence": 0.87,
+            "model_source": "claude-3",
+            "reasoning_steps": [
+                {"step": 1, "thought": "First thought", "conclusion": "First conclusion"}
+            ]
+        },
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert response.status_code == 200
+    data = response.json()
+    assert data["reasoning_type"] == "chain"
+    assert data["confidence"] == 0.87
+    assert data["model_source"] == "claude-3"
+    assert len(data["reasoning_steps"]) == 1
+
+
+@pytest.mark.asyncio
+async def test_patch_reasoning_invalid_confidence(client):
+    """Test that invalid confidence values are rejected."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "Reasoning Doc 2"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    # Invalid confidence > 1.0 - Pydantic validation returns 422
+    response = await client.patch(
+        f"/api/v1/documents/{doc_id}/reasoning",
+        json={"confidence": 1.5},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert response.status_code == 422
+    
+    # Invalid confidence < 0.0 - Pydantic validation returns 422
+    response = await client.patch(
+        f"/api/v1/documents/{doc_id}/reasoning",
+        json={"confidence": -0.1},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert response.status_code == 422
+
+
+@pytest.mark.asyncio
+async def test_get_reasoning_panel(client):
+    """Test getting reasoning panel data."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "Panel Doc"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    # Initially no reasoning
+    panel_resp = await client.get(
+        f"/api/v1/documents/{doc_id}/reasoning-panel",
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert panel_resp.status_code == 200
+    panel = panel_resp.json()
+    assert panel["has_reasoning"] is False
+    assert panel["reasoning"] is None
+    assert panel["editable"] is True
+    
+    # Add reasoning
+    await client.patch(
+        f"/api/v1/documents/{doc_id}/reasoning",
+        json={"reasoning_type": "idea", "confidence": 0.95},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    
+    # Now has reasoning
+    panel_resp = await client.get(
+        f"/api/v1/documents/{doc_id}/reasoning-panel",
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    panel = panel_resp.json()
+    assert panel["has_reasoning"] is True
+    assert panel["reasoning"]["reasoning_type"] == "idea"
+
+
+@pytest.mark.asyncio
+async def test_add_reasoning_step(client):
+    """Test adding a reasoning step."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "Steps Doc"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    # Add first step
+    step1_resp = await client.post(
+        f"/api/v1/documents/{doc_id}/reasoning-steps",
+        json={"thought": "First step thought", "conclusion": "First conclusion"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert step1_resp.status_code == 201
+    step1 = step1_resp.json()
+    assert step1["step"] == 1
+    assert step1["thought"] == "First step thought"
+    
+    # Add second step
+    step2_resp = await client.post(
+        f"/api/v1/documents/{doc_id}/reasoning-steps",
+        json={"thought": "Second step thought"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert step2_resp.status_code == 201
+    step2 = step2_resp.json()
+    assert step2["step"] == 2
+
+
+@pytest.mark.asyncio
+async def test_delete_reasoning_step(client):
+    """Test deleting a reasoning step."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "Delete Step Doc"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    # Add steps
+    await client.post(
+        f"/api/v1/documents/{doc_id}/reasoning-steps",
+        json={"thought": "Step 1"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    await client.post(
+        f"/api/v1/documents/{doc_id}/reasoning-steps",
+        json={"thought": "Step 2"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    
+    # Delete step 1
+    del_resp = await client.delete(
+        f"/api/v1/documents/{doc_id}/reasoning-steps/1",
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert del_resp.status_code == 204
+    
+    # Verify step 2 still exists
+    reasoning_resp = await client.get(
+        f"/api/v1/documents/{doc_id}/reasoning",
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    steps = reasoning_resp.json()["reasoning_steps"]
+    assert len(steps) == 1
+    assert steps[0]["step"] == 2
+
+
+@pytest.mark.asyncio
+async def test_delete_nonexistent_step(client):
+    """Test deleting a step that doesn't exist."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "No Steps Doc"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    del_resp = await client.delete(
+        f"/api/v1/documents/{doc_id}/reasoning-steps/999",
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert del_resp.status_code == 404
+
+
+# =============================================================================
+# Phase 2: TipTap Content Endpoint Tests
+# =============================================================================
+
+@pytest.mark.asyncio
+async def test_get_content_tiptap_format(client):
+    """Test getting content in TipTap format."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "TipTap Doc", "content": "Hello world"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    response = await client.get(
+        f"/api/v1/documents/{doc_id}/content",
+        params={"format": "tiptap"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert response.status_code == 200
+    data = response.json()
+    assert data["format"] == "tiptap"
+    assert "content" in data
+
+
+@pytest.mark.asyncio
+async def test_get_content_markdown_format(client):
+    """Test getting content in Markdown format."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "Markdown Doc", "content": "# Hello\n\nWorld"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    response = await client.get(
+        f"/api/v1/documents/{doc_id}/content",
+        params={"format": "markdown"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert response.status_code == 200
+    data = response.json()
+    assert data["format"] == "markdown"
+    assert "# Hello" in data["content"]
+
+
+@pytest.mark.asyncio
+async def test_put_content_tiptap(client):
+    """Test updating content with TipTap JSON."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "Update TipTap Doc"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    tiptap_content = {
+        "type": "doc",
+        "content": [
+            {
+                "type": "heading",
+                "attrs": {"level": 1},
+                "content": [{"type": "text", "text": "My Title"}]
+            },
+            {
+                "type": "paragraph",
+                "content": [{"type": "text", "text": "Hello world"}]
+            }
+        ]
+    }
+    
+    response = await client.put(
+        f"/api/v1/documents/{doc_id}/content",
+        json={"content": tiptap_content, "format": "tiptap"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    print(f"DEBUG: status={response.status_code} body={response.text}")
+    assert response.status_code == 200, f"Expected 200, got {response.status_code}: {response.text}"
+    data = response.json()
+    assert data["tiptap_content"] is not None
+    
+    # Verify markdown content was also updated
+    assert "My Title" in data["content"] or "# My Title" in data["content"]
+
+
+@pytest.mark.asyncio
+async def test_put_content_markdown(client):
+    """Test updating content with Markdown."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "Update MD Doc"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    response = await client.put(
+        f"/api/v1/documents/{doc_id}/content",
+        json={"content": "# New Title\n\nNew content here", "format": "markdown"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    assert response.status_code == 200
+    data = response.json()
+    assert data["content"] == "# New Title\n\nNew content here"
+
+
+@pytest.mark.asyncio
+async def test_document_response_includes_reasoning_fields(client):
+    """Test that DocumentResponse includes new reasoning fields."""
+    token, proj_id = await setup_project_and_get_token(client)
+    doc_resp = await client.post(
+        f"/api/v1/projects/{proj_id}/documents",
+        json={"title": "Full Doc"},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    doc_id = doc_resp.json()["id"]
+    
+    # Add reasoning
+    await client.patch(
+        f"/api/v1/documents/{doc_id}/reasoning",
+        json={"reasoning_type": "reflection", "confidence": 0.99},
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    
+    # Get document and verify fields
+    get_resp = await client.get(
+        f"/api/v1/documents/{doc_id}",
+        headers={"Authorization": f"Bearer {token}"}
+    )
+    data = get_resp.json()
+    assert "reasoning_type" in data
+    assert "confidence" in data
+    assert "reasoning_steps" in data
+    assert "model_source" in data
+    assert "tiptap_content" in data
+    assert data["reasoning_type"] == "reflection"
+    assert data["confidence"] == 0.99