Spaces:
Sleeping
Sleeping
File size: 13,194 Bytes
6252f54 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 | """
Chat endpoints — streaming EA Advisor with session persistence and DRL enrichment.
Powered by AMD MI300X + Knowledge Graph RAG.
"""
import asyncio
import json
import logging
import uuid
from typing import Annotated
from fastapi import APIRouter, Depends, Query, HTTPException
from fastapi.responses import StreamingResponse
from pydantic import BaseModel
from backend.config import Settings, get_settings
from backend.dependencies import get_neo4j_client, get_llm_client
from backend.graph.neo4j_client import Neo4jClient
from backend.graph.cypher_queries import (
ENSURE_CHAT_SESSION,
APPEND_CHAT_EXCHANGE,
GET_SESSION_MESSAGES,
GET_RECENT_CHAT_SESSIONS,
DELETE_CHAT_SESSION,
CHECK_DOMAIN_DRL_STATUS,
)
from backend.llm.client import LLMClient
log = logging.getLogger(__name__)
router = APIRouter()
CHAT_SYSTEM = """You are an Enterprise Architecture Advisor powered by AMD MI300X and Qwen-72B.
You have access to a knowledge graph containing 1,416 enterprise capabilities across 44 strategic domains,
44 governance standards, and 44 digital transformation trends.
Answer questions about enterprise architecture, digital transformation, governance, compliance, and technology strategy.
Be specific and actionable. Cite specific capabilities, governance standards, and trends from the context provided.
When relevant, mention which domain or subdomain a capability belongs to.
If asked to generate a roadmap, suggest the user use the Strategic Roadmap tab for a full structured output."""
# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------
def _classify_intent(message: str) -> str:
msg = message.lower()
if any(kw in msg for kw in ["roadmap", "transformation plan", "implementation plan"]):
return "generate_roadmap"
if any(kw in msg for kw in ["show me", "list", "what domains", "explore"]):
return "explore_domain"
return "general_qa"
async def _build_rag_context(
message: str,
neo4j: Neo4jClient,
llm: LLMClient,
) -> tuple[str, list[dict], list[str]]:
"""
Run graph RAG.
Returns (context_str, sources_list, domain_names_retrieved).
"""
from backend.agents.retriever import RetrieverAgent
retriever = RetrieverAgent(neo4j=neo4j, llm=llm)
try:
caps = await retriever.retrieve(org_type=message[:80], goals=[message], sectors=[])
except Exception as exc:
log.warning(f"RAG retrieval failed: {exc}")
caps = []
sources: list[dict] = []
context_lines: list[str] = []
domain_names: list[str] = []
for c in caps[:6]:
cap = c.capability
std = c.standard or {}
trend = c.trend or {}
name = cap.get("name", "")
domain = c.domain.get("name", "")
outcomes = (cap.get("business_outcomes") or [])[:2]
context_lines.append(
f"- Capability: {name} (Domain: {domain})\n"
f" Outcomes: {'; '.join(outcomes)}\n"
f" Governance: {std.get('name', '')}\n"
f" Trend: {trend.get('name', '')}"
)
sources.append({
"name": name,
"domain": domain,
"standard": std.get("name", ""),
"trend": trend.get("name", ""),
})
if domain and domain not in domain_names:
domain_names.append(domain)
context = "\n".join(context_lines) if context_lines else "No specific capabilities matched."
return context, sources, domain_names
def _build_messages(history: list[dict], user_message: str, context: str) -> list[dict]:
messages = []
for msg in history[-8:]:
if msg.get("role") in ("user", "assistant") and msg.get("content"):
# Skip the initial welcome assistant message to avoid invalid conversation order
if msg["role"] == "assistant" and "Enterprise Architecture Advisor" in msg["content"]:
continue
messages.append({"role": msg["role"], "content": msg["content"]})
messages.append({
"role": "user",
"content": f"KNOWLEDGE GRAPH CONTEXT:\n{context}\n\nQUESTION: {user_message}",
})
return messages
def _get_untrained_domains(domain_names: list[str], neo4j: Neo4jClient) -> list[str]:
"""Return subset of domain_names that have not yet been DRL-trained."""
if not domain_names:
return []
try:
rows = neo4j.run_query(CHECK_DOMAIN_DRL_STATUS, domain_names=domain_names)
return [r["name"] for r in rows if not r.get("trained")]
except Exception as exc:
log.warning(f"DRL status check failed: {exc}")
return []
async def _trigger_drl_enrichment(domain_names: list[str]) -> str:
"""Fire-and-forget background DRL training for untrained domains."""
run_id = uuid.uuid4().hex[:10]
async def _run():
try:
import os
from dotenv import load_dotenv
load_dotenv()
from neo4j import GraphDatabase
from pipeline.train_on_graph import GraphTrainer
driver = GraphDatabase.driver(
os.getenv("NEO4J_URI"),
auth=(os.getenv("NEO4J_USERNAME"), os.getenv("NEO4J_PASSWORD")),
)
try:
trainer = GraphTrainer(driver, os.getenv("NEO4J_DATABASE", "neo4j"))
loop = asyncio.get_event_loop()
for domain in domain_names:
await loop.run_in_executor(
None,
lambda d=domain: trainer.run(episodes_per_domain=50, target_domain=d),
)
log.info(f"DRL enrichment {run_id} complete for: {domain_names}")
finally:
driver.close()
except Exception as exc:
log.warning(f"DRL enrichment {run_id} failed: {exc}")
asyncio.create_task(_run())
return run_id
def _store_exchange(
neo4j: Neo4jClient,
session_id: str,
user_content: str,
asst_content: str,
sources: list[dict],
title: str,
) -> None:
"""Persist a user↔assistant exchange to Neo4j."""
try:
neo4j.run_query(ENSURE_CHAT_SESSION, session_id=session_id, title=title)
neo4j.run_query(
APPEND_CHAT_EXCHANGE,
session_id=session_id,
user_msg_id=uuid.uuid4().hex,
user_content=user_content,
asst_msg_id=uuid.uuid4().hex,
asst_content=asst_content,
sources_json=json.dumps(sources),
)
except Exception as exc:
log.warning(f"Failed to store chat exchange: {exc}")
# ---------------------------------------------------------------------------
# Session management endpoints
# ---------------------------------------------------------------------------
@router.post("/chat/sessions")
async def create_session(
neo4j: Annotated[Neo4jClient, Depends(get_neo4j_client)],
session_id: str | None = None,
title: str = "",
):
"""Create or touch a chat session. Returns session_id."""
sid = session_id or uuid.uuid4().hex
try:
neo4j.run_query(ENSURE_CHAT_SESSION, session_id=sid, title=title or "New Conversation")
except Exception as exc:
log.warning(f"Session create failed: {exc}")
return {"session_id": sid}
@router.get("/chat/sessions")
async def list_sessions(neo4j: Annotated[Neo4jClient, Depends(get_neo4j_client)]):
"""Return the 15 most recent chat sessions."""
try:
rows = neo4j.run_query(GET_RECENT_CHAT_SESSIONS)
return [
{
"session_id": r["session_id"],
"title": r.get("title", "Untitled"),
"last_active": r.get("last_active", ""),
"message_count": r.get("message_count", 0),
}
for r in rows
]
except Exception as exc:
log.warning(f"List sessions failed: {exc}")
return []
@router.get("/chat/sessions/{session_id}/messages")
async def get_session(
session_id: str,
neo4j: Annotated[Neo4jClient, Depends(get_neo4j_client)],
):
"""Return all messages for a session."""
try:
rows = neo4j.run_query(GET_SESSION_MESSAGES, session_id=session_id)
messages = []
for r in rows:
srcs = []
try:
srcs = json.loads(r.get("sources_json", "[]"))
except Exception:
pass
messages.append({
"role": r["role"],
"content": r["content"],
"sources": srcs,
"created_at": r.get("created_at", ""),
})
return {"session_id": session_id, "messages": messages}
except Exception as exc:
raise HTTPException(status_code=500, detail=str(exc))
@router.delete("/chat/sessions/{session_id}")
async def delete_session(
session_id: str,
neo4j: Annotated[Neo4jClient, Depends(get_neo4j_client)],
):
"""Delete a chat session and all its messages."""
try:
neo4j.run_query(DELETE_CHAT_SESSION, session_id=session_id)
return {"deleted": session_id}
except Exception as exc:
raise HTTPException(status_code=500, detail=str(exc))
# ---------------------------------------------------------------------------
# Chat endpoints
# ---------------------------------------------------------------------------
class ChatRequest(BaseModel):
message: str
history: list[dict] = []
domain_filter: str | None = None
session_id: str | None = None
@router.post("/chat")
async def chat(
request: ChatRequest,
settings: Annotated[Settings, Depends(get_settings)],
neo4j: Annotated[Neo4jClient, Depends(get_neo4j_client)],
llm: Annotated[LLMClient, Depends(get_llm_client)],
):
intent = _classify_intent(request.message)
context, sources, domain_names = await _build_rag_context(request.message, neo4j, llm)
messages = _build_messages(request.history, request.message, context)
reply = await llm.chat(
messages=messages,
system=CHAT_SYSTEM,
max_tokens=settings.llm_max_tokens,
temperature=0.4,
)
# Persist if session provided
if request.session_id:
title = request.message[:60]
_store_exchange(neo4j, request.session_id, request.message, reply, sources, title)
# Check DRL enrichment
untrained = _get_untrained_domains(domain_names, neo4j)
enrich_run_id = None
if untrained:
enrich_run_id = await _trigger_drl_enrichment(untrained)
log.info(f"DRL enrichment triggered for: {untrained} (run {enrich_run_id})")
return {
"reply": reply,
"sources": sources,
"suggested_action": intent if intent != "general_qa" else None,
"enrich_triggered": bool(untrained),
"enrich_domains": untrained,
}
@router.get("/chat/stream")
async def chat_stream(
message: str = Query(...),
history: str = Query(default="[]"),
session_id: str = Query(default=""),
settings: Annotated[Settings, Depends(get_settings)] = None,
neo4j: Annotated[Neo4jClient, Depends(get_neo4j_client)] = None,
llm: Annotated[LLMClient, Depends(get_llm_client)] = None,
):
history_list: list[dict] = []
try:
history_list = json.loads(history)
except Exception:
pass
context, sources, domain_names = await _build_rag_context(message, neo4j, llm)
messages = _build_messages(history_list, message, context)
# Check untrained domains before streaming
untrained = _get_untrained_domains(domain_names, neo4j)
async def generate():
full_chunks: list[str] = []
try:
async for chunk in llm.chat_stream(
messages=messages,
system=CHAT_SYSTEM,
max_tokens=settings.llm_max_tokens,
):
full_chunks.append(chunk)
yield f"data: {json.dumps({'text': chunk})}\n\n"
except Exception as exc:
log.error(f"Stream error: {exc}")
yield f"data: {json.dumps({'text': f' [Stream error: {exc}]'})}\n\n"
finally:
full_response = "".join(full_chunks)
# Persist exchange
if session_id:
title = message[:60]
_store_exchange(neo4j, session_id, message, full_response, sources, title)
# Trigger DRL enrichment in background
enrich_run_id = None
if untrained:
try:
enrich_run_id = await _trigger_drl_enrichment(untrained)
log.info(f"DRL enrichment triggered for {untrained} (run {enrich_run_id})")
except Exception as exc:
log.warning(f"DRL trigger failed: {exc}")
yield f"data: {json.dumps({'sources': sources, 'enrich_triggered': bool(untrained), 'enrich_domains': untrained, 'enrich_run_id': enrich_run_id, 'done': True})}\n\n"
return StreamingResponse(
generate(),
media_type="text/event-stream",
headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"},
)
|