Merge pull request #724 from langflow-ai/sdk-chat-fix
sdk chat endpoint fix
This commit is contained in:
Sebastián Estévez
GitHub
commit
a252c94312
1 changed files with 44 additions and 116 deletions
|
|
@ -2,7 +2,7 @@
|
|||
Public API v1 Chat endpoint.
|
||||
|
||||
Provides chat functionality with streaming support and conversation history.
|
||||
Uses API key authentication.
|
||||
Uses API key authentication. Routes through Langflow (chat_service.langflow_chat).
|
||||
"""
|
||||
import json
|
||||
from starlette.requests import Request
|
||||
|
|
@ -15,25 +15,18 @@ logger = get_logger(__name__)
|
|||
|
||||
async def _transform_stream_to_sse(raw_stream, chat_id_container: dict):
|
||||
"""
|
||||
Transform the raw internal streaming format to clean SSE events.
|
||||
Transform the raw Langflow streaming format to clean SSE events for v1 API.
|
||||
|
||||
Yields SSE events in the format:
|
||||
event: content
|
||||
data: {"type": "content", "delta": "..."}
|
||||
|
||||
event: sources
|
||||
data: {"type": "sources", "sources": [...]}
|
||||
|
||||
event: done
|
||||
data: {"type": "done", "chat_id": "..."}
|
||||
"""
|
||||
full_text = ""
|
||||
sources = []
|
||||
chat_id = None
|
||||
|
||||
async for chunk in raw_stream:
|
||||
try:
|
||||
# Decode the chunk
|
||||
if isinstance(chunk, bytes):
|
||||
chunk_str = chunk.decode("utf-8").strip()
|
||||
else:
|
||||
|
|
@ -42,131 +35,76 @@ async def _transform_stream_to_sse(raw_stream, chat_id_container: dict):
|
|||
if not chunk_str:
|
||||
continue
|
||||
|
||||
# Parse the JSON chunk
|
||||
chunk_data = json.loads(chunk_str)
|
||||
|
||||
# Extract text delta
|
||||
delta_text = None
|
||||
# Extract text from various possible formats
|
||||
delta_text = ""
|
||||
|
||||
# Format 1: delta.content (OpenAI-style)
|
||||
if "delta" in chunk_data:
|
||||
delta = chunk_data["delta"]
|
||||
if isinstance(delta, dict):
|
||||
delta_text = delta.get("content") or delta.get("text") or ""
|
||||
delta_text = delta.get("content", "") or delta.get("text", "")
|
||||
elif isinstance(delta, str):
|
||||
delta_text = delta
|
||||
|
||||
if "output_text" in chunk_data and chunk_data["output_text"]:
|
||||
# Format 2: output_text (Langflow-style)
|
||||
if not delta_text and chunk_data.get("output_text"):
|
||||
delta_text = chunk_data["output_text"]
|
||||
|
||||
# Yield content event if we have text
|
||||
# Format 3: text field directly
|
||||
if not delta_text and chunk_data.get("text"):
|
||||
delta_text = chunk_data["text"]
|
||||
|
||||
# Format 4: content field directly
|
||||
if not delta_text and chunk_data.get("content"):
|
||||
delta_text = chunk_data["content"]
|
||||
|
||||
if delta_text:
|
||||
full_text += delta_text
|
||||
event = {"type": "content", "delta": delta_text}
|
||||
yield f"event: content\ndata: {json.dumps(event)}\n\n"
|
||||
yield f"data: {json.dumps({'type': 'content', 'delta': delta_text})}\n\n"
|
||||
|
||||
# Extract chat_id/response_id
|
||||
if "id" in chunk_data and chunk_data["id"]:
|
||||
chat_id = chunk_data["id"]
|
||||
elif "response_id" in chunk_data and chunk_data["response_id"]:
|
||||
chat_id = chunk_data["response_id"]
|
||||
|
||||
# Extract sources from tool call results
|
||||
if "item" in chunk_data and isinstance(chunk_data["item"], dict):
|
||||
item = chunk_data["item"]
|
||||
if item.get("type") in ("retrieval_call", "tool_call", "function_call"):
|
||||
results = item.get("results", [])
|
||||
if results:
|
||||
for result in results:
|
||||
if isinstance(result, dict):
|
||||
source = {
|
||||
"filename": result.get("filename", result.get("title", "Unknown")),
|
||||
"text": result.get("text", result.get("content", "")),
|
||||
"score": result.get("score", 0),
|
||||
"page": result.get("page"),
|
||||
}
|
||||
sources.append(source)
|
||||
# Extract chat_id/response_id from various fields
|
||||
if not chat_id:
|
||||
chat_id = chunk_data.get("id") or chunk_data.get("response_id")
|
||||
|
||||
except json.JSONDecodeError:
|
||||
# Not JSON, might be raw text
|
||||
if chunk_str and not chunk_str.startswith("{"):
|
||||
event = {"type": "content", "delta": chunk_str}
|
||||
yield f"event: content\ndata: {json.dumps(event)}\n\n"
|
||||
# Raw text without JSON wrapper
|
||||
if chunk_str:
|
||||
yield f"data: {json.dumps({'type': 'content', 'delta': chunk_str})}\n\n"
|
||||
full_text += chunk_str
|
||||
except Exception as e:
|
||||
logger.warning("Error processing stream chunk", error=str(e))
|
||||
continue
|
||||
logger.warning("Error processing stream chunk", error=str(e), chunk=chunk_str[:100] if chunk_str else "")
|
||||
|
||||
# Yield sources event if we have any
|
||||
if sources:
|
||||
event = {"type": "sources", "sources": sources}
|
||||
yield f"event: sources\ndata: {json.dumps(event)}\n\n"
|
||||
|
||||
# Yield done event with chat_id
|
||||
event = {"type": "done", "chat_id": chat_id}
|
||||
yield f"event: done\ndata: {json.dumps(event)}\n\n"
|
||||
|
||||
# Store chat_id for caller
|
||||
yield f"data: {json.dumps({'type': 'done', 'chat_id': chat_id})}\n\n"
|
||||
chat_id_container["chat_id"] = chat_id
|
||||
|
||||
|
||||
async def chat_create_endpoint(request: Request, chat_service, session_manager):
|
||||
"""
|
||||
Send a chat message.
|
||||
Send a chat message via Langflow.
|
||||
|
||||
POST /v1/chat
|
||||
|
||||
Request body:
|
||||
{
|
||||
"message": "What is RAG?",
|
||||
"stream": false, // optional, default false
|
||||
"chat_id": "...", // optional, to continue conversation
|
||||
"filters": {...}, // optional
|
||||
"limit": 10, // optional
|
||||
"score_threshold": 0.5 // optional
|
||||
}
|
||||
|
||||
Non-streaming response:
|
||||
{
|
||||
"response": "RAG stands for...",
|
||||
"chat_id": "chat_xyz789",
|
||||
"sources": [...]
|
||||
}
|
||||
|
||||
Streaming response (SSE):
|
||||
event: content
|
||||
data: {"type": "content", "delta": "RAG stands for"}
|
||||
|
||||
event: sources
|
||||
data: {"type": "sources", "sources": [...]}
|
||||
|
||||
event: done
|
||||
data: {"type": "done", "chat_id": "chat_xyz789"}
|
||||
"""
|
||||
try:
|
||||
data = await request.json()
|
||||
except Exception:
|
||||
return JSONResponse(
|
||||
{"error": "Invalid JSON in request body"},
|
||||
status_code=400,
|
||||
)
|
||||
return JSONResponse({"error": "Invalid JSON in request body"}, status_code=400)
|
||||
|
||||
message = data.get("message", "").strip()
|
||||
if not message:
|
||||
return JSONResponse(
|
||||
{"error": "Message is required"},
|
||||
status_code=400,
|
||||
)
|
||||
return JSONResponse({"error": "Message is required"}, status_code=400)
|
||||
|
||||
stream = data.get("stream", False)
|
||||
chat_id = data.get("chat_id") # For conversation continuation
|
||||
chat_id = data.get("chat_id")
|
||||
filters = data.get("filters")
|
||||
limit = data.get("limit", 10)
|
||||
score_threshold = data.get("score_threshold", 0)
|
||||
filter_id = data.get("filter_id")
|
||||
|
||||
user = request.state.user
|
||||
user_id = user.user_id
|
||||
|
||||
# Note: API key auth doesn't have JWT, so we pass None
|
||||
jwt_token = None
|
||||
jwt_token = session_manager.get_effective_jwt_token(user_id, None)
|
||||
|
||||
# Set context variables for search tool
|
||||
if filters:
|
||||
|
|
@ -176,45 +114,35 @@ async def chat_create_endpoint(request: Request, chat_service, session_manager):
|
|||
set_auth_context(user_id, jwt_token)
|
||||
|
||||
if stream:
|
||||
# Streaming response
|
||||
raw_stream = await chat_service.chat(
|
||||
raw_stream = await chat_service.langflow_chat(
|
||||
prompt=message,
|
||||
user_id=user_id,
|
||||
jwt_token=jwt_token,
|
||||
previous_response_id=chat_id,
|
||||
stream=True,
|
||||
filter_id=filter_id,
|
||||
)
|
||||
|
||||
chat_id_container = {}
|
||||
|
||||
return StreamingResponse(
|
||||
_transform_stream_to_sse(raw_stream, chat_id_container),
|
||||
media_type="text/event-stream",
|
||||
headers={
|
||||
"Cache-Control": "no-cache",
|
||||
"Connection": "keep-alive",
|
||||
"X-Accel-Buffering": "no",
|
||||
},
|
||||
headers={"Cache-Control": "no-cache", "Connection": "keep-alive", "X-Accel-Buffering": "no"},
|
||||
)
|
||||
else:
|
||||
# Non-streaming response
|
||||
result = await chat_service.chat(
|
||||
result = await chat_service.langflow_chat(
|
||||
prompt=message,
|
||||
user_id=user_id,
|
||||
jwt_token=jwt_token,
|
||||
previous_response_id=chat_id,
|
||||
stream=False,
|
||||
filter_id=filter_id,
|
||||
)
|
||||
|
||||
# Transform response to public API format
|
||||
# Internal format: {"response": "...", "response_id": "..."}
|
||||
response_data = {
|
||||
# Transform response_id to chat_id for v1 API format
|
||||
return JSONResponse({
|
||||
"response": result.get("response", ""),
|
||||
"chat_id": result.get("response_id"),
|
||||
"sources": result.get("sources", []),
|
||||
}
|
||||
|
||||
return JSONResponse(response_data)
|
||||
})
|
||||
|
||||
|
||||
async def chat_list_endpoint(request: Request, chat_service, session_manager):
|
||||
|
|
@ -240,8 +168,8 @@ async def chat_list_endpoint(request: Request, chat_service, session_manager):
|
|||
user_id = user.user_id
|
||||
|
||||
try:
|
||||
# Get chat history
|
||||
history = await chat_service.get_chat_history(user_id)
|
||||
# Get Langflow chat history (since v1 routes through Langflow)
|
||||
history = await chat_service.get_langflow_history(user_id)
|
||||
|
||||
# Transform to public API format
|
||||
conversations = []
|
||||
|
|
@ -293,8 +221,8 @@ async def chat_get_endpoint(request: Request, chat_service, session_manager):
|
|||
)
|
||||
|
||||
try:
|
||||
# Get chat history and find the specific conversation
|
||||
history = await chat_service.get_chat_history(user_id)
|
||||
# Get Langflow chat history and find the specific conversation
|
||||
history = await chat_service.get_langflow_history(user_id)
|
||||
|
||||
conversation = None
|
||||
for conv in history.get("conversations", []):
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue