feat: add confluence/slack search tools, chat history, cloud Qdrant support, sync trigger fixes

.dockerignore

@@ -0,0 +1,14 @@
+.env
+.venv
+venv
+__pycache__
+*.pyc
+*.pyo
+.git
+.gitignore
+frontend/node_modules
+frontend/src
+frontend/.env
+data/bm25_index.pkl
+*.log
+.DS_Store

.github/workflows/keep-alive.yml

@@ -0,0 +1,12 @@
+name: Keep HF Space Alive
+on:
+  schedule:
+    - cron: '*/30 * * * *'
+  workflow_dispatch:
+
+jobs:
+  ping:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Ping space
+        run: curl -sf ${{ secrets.HF_SPACE_URL }}/health || true

.gitignore

@@ -38,3 +38,6 @@ Thumbs.db
 celerybeat-schedule
 celerybeat.pid
 .env
+frontend/node_modules/
+frontend/package-lock.json
+frontend/tsconfig.tsbuildinfo

Dockerfile

@@ -0,0 +1,26 @@
+FROM python:3.11-slim
+
+# System deps
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    build-essential curl git \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /app
+
+# Install Python deps first (layer cache)
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Pre-download the BGE-M3 model so cold starts are fast
+RUN python3 -c "from FlagEmbedding import BGEM3FlagModel; BGEM3FlagModel('BAAI/bge-m3', use_fp16=False)"
+
+# Download spacy model
+RUN python3 -m spacy download en_core_web_sm
+
+# Copy app code
+COPY . .
+
+# HuggingFace Spaces requires port 7860
+EXPOSE 7860
+
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -1,12 +1,3 @@
----
-title: GodSpeed
-emoji: 🚀
-colorFrom: red
-colorTo: yellow
-sdk: docker
-pinned: false
----
-
 # Godspeed
 
 Enterprise Knowledge Copilot is a fully open-source, locally-compliant, agentic RAG platform that unifies internal and live external knowledge into a single cited, validated answer engine — purpose-built for IT enterprises operating under GDPR and India's DPDP Act.

agent/agents/_gemini.py

@@ -1,4 +1,4 @@
-"""Shared Gemini call helper with exponential-backoff retry."""
+"""LLM call helper with exponential-backoff retry. Uses OpenAI when key is set, else Gemini."""
 
 from __future__ import annotations
 
@@ -7,7 +7,6 @@ import json
 import logging
 from typing import Any
 
-from langchain_google_genai import ChatGoogleGenerativeAI
 from langchain_core.messages import HumanMessage, SystemMessage
 
 from agent.config import settings
@@ -15,16 +14,33 @@ from agent.config import settings
 logger = logging.getLogger(__name__)
 
 
+def _make_llm(model_name: str, streaming: bool = False):
+    openai_key = getattr(settings, "openai_api_key", "")
+    if openai_key:
+        from langchain_openai import ChatOpenAI
+        # Map Gemini model names to OpenAI equivalents
+        oai_model = "gpt-4o-mini"
+        return ChatOpenAI(
+            model=oai_model,
+            api_key=openai_key,
+            temperature=0.0,
+            streaming=streaming,
+        )
+    from langchain_google_genai import ChatGoogleGenerativeAI
+    return ChatGoogleGenerativeAI(
+        model=model_name,
+        google_api_key=settings.google_api_key,
+        temperature=0.0,
+        streaming=streaming,
+    )
+
+
 async def call_gemini_text(
     model_name: str,
     system_prompt: str,
     user_message: str,
 ) -> str:
-    llm = ChatGoogleGenerativeAI(
-        model=model_name,
-        google_api_key=settings.google_api_key,
-        temperature=0.0,
-    )
+    llm = _make_llm(model_name)
     messages = [SystemMessage(content=system_prompt), HumanMessage(content=user_message)]
 
     for attempt in range(settings.gemini_max_retries):
@@ -33,10 +49,10 @@ async def call_gemini_text(
             return response.content
         except Exception as exc:
             if attempt == settings.gemini_max_retries - 1:
-                logger.error("Gemini call failed after %d retries: %s", settings.gemini_max_retries, exc)
+                logger.error("LLM call failed after %d retries: %s", settings.gemini_max_retries, exc)
                 raise
             delay = settings.gemini_retry_base_delay * (2 ** attempt)
-            logger.warning("Gemini call attempt %d failed (%s); retrying in %.1fs", attempt + 1, exc, delay)
+            logger.warning("LLM call attempt %d failed (%s); retrying in %.1fs", attempt + 1, exc, delay)
             await asyncio.sleep(delay)
 
     raise RuntimeError("Unreachable")
@@ -48,7 +64,6 @@ async def call_gemini_json(
     user_message: str,
 ) -> dict[str, Any]:
     raw = await call_gemini_text(model_name, system_prompt, user_message)
-    # Strip markdown code fences if model ignores the instruction
     cleaned = raw.strip()
     if cleaned.startswith("```"):
         cleaned = cleaned.split("\n", 1)[-1]
@@ -57,7 +72,7 @@ async def call_gemini_json(
     try:
         return json.loads(cleaned)
     except json.JSONDecodeError as exc:
-        logger.error("Failed to parse Gemini JSON response: %s\nRaw: %s", exc, raw[:500])
+        logger.error("Failed to parse LLM JSON response: %s\nRaw: %s", exc, raw[:500])
         raise
 
 
@@ -66,12 +81,7 @@ async def stream_gemini_text(
     system_prompt: str,
     user_message: str,
 ):
-    llm = ChatGoogleGenerativeAI(
-        model=model_name,
-        google_api_key=settings.google_api_key,
-        temperature=0.0,
-        streaming=True,
-    )
+    llm = _make_llm(model_name, streaming=True)
     messages = [SystemMessage(content=system_prompt), HumanMessage(content=user_message)]
 
     for attempt in range(settings.gemini_max_retries):
@@ -81,8 +91,8 @@ async def stream_gemini_text(
             return
         except Exception as exc:
             if attempt == settings.gemini_max_retries - 1:
-                logger.error("Gemini stream failed after %d retries: %s", settings.gemini_max_retries, exc)
+                logger.error("LLM stream failed after %d retries: %s", settings.gemini_max_retries, exc)
                 raise
             delay = settings.gemini_retry_base_delay * (2 ** attempt)
-            logger.warning("Gemini stream attempt %d failed (%s); retrying in %.1fs", attempt + 1, exc, delay)
+            logger.warning("LLM stream attempt %d failed (%s); retrying in %.1fs", attempt + 1, exc, delay)
             await asyncio.sleep(delay)

agent/api.py

@@ -31,12 +31,17 @@ async def _store_query_event(
     agent_results: dict | None = None,
     guardrail_score: float | None = None,
     escalated: bool = False,
+    answer_text: str = "",
 ) -> None:
     """Persist query event to Redis for analytics and workspace history."""
     try:
         import redis.asyncio as aioredis
         from src.config import settings
 
+        brief = answer_text[:500].rstrip() if answer_text else ""
+        if answer_text and len(answer_text) > 500:
+            brief += "…"
+
         event = {
             "id":           str(uuid4()),
             "query":        query_input.query,
@@ -45,7 +50,7 @@ async def _store_query_event(
             "created_at":   datetime.utcnow().isoformat(),
             "success":      success,
             "duration_ms":  duration_ms,
-            "answer_brief": "",
+            "answer_brief": brief,
             # Per-agent retrieval metrics — populated from final graph state
             "agents": {
                 agent: {
@@ -122,6 +127,7 @@ async def _event_generator(
                     agent_results=final_state.get("agent_results", {}),
                     guardrail_score=final_state.get("guardrail_score"),
                     escalated=final_state.get("escalate", False),
+                    answer_text=final_state.get("final_answer") or "",
                 )
             except Exception as exc:
                 logger.exception(

agent/config.py

@@ -10,14 +10,17 @@ class Settings(BaseSettings):
     )
 
     google_api_key: str = ""
+    openai_api_key: str = ""
 
-    planner_model: str = "gemini-2.5-pro"
-    synthesiser_model: str = "gemini-2.5-pro"
-    summariser_model: str = "gemini-2.5-flash"
-    guardrail_model: str = "gemini-2.5-flash"
+    planner_model: str = "gemini-2.0-flash"
+    synthesiser_model: str = "gemini-2.0-flash"
+    summariser_model: str = "gemini-2.0-flash"
+    guardrail_model: str = "gemini-2.0-flash"
 
     qdrant_host: str = "localhost"
     qdrant_port: int = 6333
+    qdrant_url: str = ""        # overrides host+port when set (Qdrant Cloud)
+    qdrant_api_key: str = ""    # required for Qdrant Cloud
     qdrant_collection: str = "knowledge_base"
     qdrant_dense_vector_name: str = "dense"
     qdrant_sparse_vector_name: str = "sparse"
@@ -33,7 +36,7 @@ class Settings(BaseSettings):
     rrf_top_k: int = 50
     final_top_k: int = 5
     reranker_high_threshold: float = 0.6
-    reranker_medium_threshold: float = 0.4
+    reranker_medium_threshold: float = 0.3
     live_docs_confidence_threshold: float = 0.5
 
     gemini_max_retries: int = 3
@@ -49,8 +52,14 @@ class Settings(BaseSettings):
     # NL-to-SQL tool — direct PostgreSQL connection string.
     # e.g. postgresql://postgres:password@db.yourproject.supabase.co:5432/postgres
     # Leave empty to disable the sql_query agent gracefully.
+    # Reads DATABASE_URL or PG_DSN from environment.
     database_url: str = ""
+    pg_dsn: str = ""
     sql_max_rows: int = 20
 
+    @property
+    def effective_database_url(self) -> str:
+        return self.database_url or self.pg_dsn
+
 
 settings = Settings()

agent/graph.py

@@ -12,8 +12,10 @@ from agent.agents.guardrail import run_guardrail
 from agent.agents.planner import run_planner
 from agent.agents.synthesiser import stream_synthesis
 from agent.models import AgentResult, KnowledgeGraphState, RetrievedChunk
+from agent.tools.confluence_search import run_confluence_search
 from agent.tools.doc_search import compute_retrieval_confidence, run_doc_search
 from agent.tools.live_docs import run_live_docs
+from agent.tools.slack_search import run_slack_search
 from agent.tools.sql_query import run_sql_query
 from agent.tools.ticket_lookup import run_ticket_lookup
 
@@ -101,6 +103,56 @@ async def ticket_lookup_node(state: KnowledgeGraphState) -> dict:
     return {"agent_results": {**state.agent_results, "ticket_lookup": result}}
 
 
+async def confluence_search_node(state: KnowledgeGraphState) -> dict:
+    queue = state.sse_queue
+    await _push_event(queue, "agent_started", {"agent": "confluence_search"})
+
+    task_input = _find_task_input(state, "confluence_search") or state.query_input.query
+    chunks: list[RetrievedChunk] = []
+    error: str | None = None
+
+    try:
+        chunks = await run_confluence_search(task_input, state.query_input.team_id)
+    except Exception as exc:
+        logger.exception("confluence_search_node error")
+        error = str(exc)
+
+    confidence = compute_retrieval_confidence(chunks)
+    result = AgentResult(
+        agent="confluence_search",
+        chunks=chunks,
+        retrieval_confidence=confidence,
+        error=error,
+    )
+    await _push_event(queue, "agent_done", {"agent": "confluence_search", "retrieval_confidence": confidence})
+    return {"agent_results": {**state.agent_results, "confluence_search": result}}
+
+
+async def slack_search_node(state: KnowledgeGraphState) -> dict:
+    queue = state.sse_queue
+    await _push_event(queue, "agent_started", {"agent": "slack_search"})
+
+    task_input = _find_task_input(state, "slack_search") or state.query_input.query
+    chunks: list[RetrievedChunk] = []
+    error: str | None = None
+
+    try:
+        chunks = await run_slack_search(task_input, state.query_input.team_id)
+    except Exception as exc:
+        logger.exception("slack_search_node error")
+        error = str(exc)
+
+    confidence = compute_retrieval_confidence(chunks)
+    result = AgentResult(
+        agent="slack_search",
+        chunks=chunks,
+        retrieval_confidence=confidence,
+        error=error,
+    )
+    await _push_event(queue, "agent_done", {"agent": "slack_search", "retrieval_confidence": confidence})
+    return {"agent_results": {**state.agent_results, "slack_search": result}}
+
+
 async def live_docs_node(state: KnowledgeGraphState) -> dict:
     queue = state.sse_queue
     await _push_event(queue, "agent_started", {"agent": "live_docs"})
@@ -246,6 +298,8 @@ def build_graph() -> Any:
     builder.add_node("planner_node", planner_node)
     builder.add_node("doc_search_node", doc_search_node)
     builder.add_node("ticket_lookup_node", ticket_lookup_node)
+    builder.add_node("confluence_search_node", confluence_search_node)
+    builder.add_node("slack_search_node", slack_search_node)
     builder.add_node("live_docs_node", live_docs_node)
     builder.add_node("sql_query_node", sql_query_node)
     builder.add_node("join_node", join_node)
@@ -260,6 +314,8 @@ def build_graph() -> Any:
         {
             "doc_search_node": "doc_search_node",
             "ticket_lookup_node": "ticket_lookup_node",
+            "confluence_search_node": "confluence_search_node",
+            "slack_search_node": "slack_search_node",
             "live_docs_node": "live_docs_node",
             "sql_query_node": "sql_query_node",
             "summariser_node": "synthesiser_node",
@@ -271,6 +327,8 @@ def build_graph() -> Any:
     # incoming edge to fire before executing join_node (fan-in).
     builder.add_edge("doc_search_node", "join_node")
     builder.add_edge("ticket_lookup_node", "join_node")
+    builder.add_edge("confluence_search_node", "join_node")
+    builder.add_edge("slack_search_node", "join_node")
     builder.add_edge("live_docs_node", "join_node")
     builder.add_edge("sql_query_node", "join_node")
 

agent/models.py

@@ -16,7 +16,7 @@ class QueryInput(BaseModel):
 
 
 class AgentTask(BaseModel):
-    agent: Literal["doc_search", "ticket_lookup", "live_docs", "summariser", "sql_query"]
+    agent: Literal["doc_search", "ticket_lookup", "confluence_search", "slack_search", "live_docs", "summariser", "sql_query"]
     input: str
     depends_on: list[str] = Field(default_factory=list)
 

agent/prompts.py

@@ -3,19 +3,23 @@ PLANNER_SYSTEM_PROMPT = """You are a planning agent for an Enterprise Knowledge
 Given a user query, decide which retrieval agents are needed and in what order.
 
 Available agents:
-- doc_search: Searches the internal knowledge base (Qdrant vector DB). Use for product docs, runbooks, internal wikis, architecture docs.
-- ticket_lookup: Searches Jira tickets. Use when query mentions bugs, issues, tickets, sprints, or task tracking.
+- doc_search: Searches the internal knowledge base (Qdrant vector DB). Use for general product docs, runbooks, architecture docs, GitHub code, and any knowledge not specific to Confluence, Jira, or Slack.
+- ticket_lookup: Searches Jira tickets. Use when query mentions bugs, issues, tickets, sprints, task tracking, or specific ticket IDs (e.g. KAN-7).
+- confluence_search: Searches Confluence pages. Use when query explicitly mentions Confluence, internal wiki pages, design docs, meeting notes, or space/page content.
+- slack_search: Searches Slack messages live. Use when query mentions Slack, team conversations, channel discussions, or asks what was discussed or said about a topic.
 - live_docs: Fetches live web content via Firecrawl/Tavily. Use ONLY when the query mentions a specific external library, framework, or third-party tool where internal docs are insufficient.
 - summariser: Summarises a large set of retrieved chunks. Use ONLY when more than 10 chunks are expected.
 - sql_query: Queries the internal Supabase database with SQL. Use when the query asks about counts, stats, aggregations, ingestion status, document lists, or any structured/numeric data (e.g. "how many documents", "failed jobs", "which source types", "ingestion stats").
 
 Rules:
-1. doc_search and ticket_lookup should run in parallel when both are needed — set depends_on: [] for both.
-2. live_docs only runs if you expect doc_search confidence will be low OR the query names a specific external library/framework. Set depends_on: ["doc_search"] to run after.
-3. summariser only runs after doc_search. Set depends_on: ["doc_search"].
-4. Do NOT include agents that are not needed for this query.
-5. Rephrase the input for each agent to be focused and specific to what that agent can retrieve.
-6. sql_query runs independently (depends_on: []). Use it when the query is about structured or aggregated data rather than semantic knowledge retrieval. It can run in parallel with doc_search.
+1. doc_search, ticket_lookup, confluence_search, and slack_search can all run in parallel — set depends_on: [] for each.
+2. Use confluence_search instead of (or in addition to) doc_search when the query is clearly about Confluence content.
+3. Use slack_search when the query is clearly about Slack conversations.
+4. live_docs only runs if you expect doc_search confidence will be low OR the query names a specific external library/framework. Set depends_on: ["doc_search"] to run after.
+5. summariser only runs after doc_search. Set depends_on: ["doc_search"].
+6. Do NOT include agents that are not needed for this query.
+7. Rephrase the input for each agent to be focused and specific to what that agent can retrieve.
+8. sql_query runs independently (depends_on: []). Use it when the query is about structured or aggregated data rather than semantic knowledge retrieval. It can run in parallel with doc_search.
 
 Return ONLY valid JSON matching this exact schema. No preamble. No markdown code fences. No explanation outside the JSON.
 

agent/tools/confluence_search.py

@@ -0,0 +1,67 @@
+"""Confluence page search — retrieves ingested Confluence chunks from Qdrant."""
+
+from __future__ import annotations
+
+import logging
+
+from agent.config import settings
+from agent.models import RetrievedChunk
+
+logger = logging.getLogger(__name__)
+
+
+async def run_confluence_search(query: str, team_id: str) -> list[RetrievedChunk]:
+    """Search Qdrant for Confluence chunks matching the query."""
+    from qdrant_client import AsyncQdrantClient
+    from qdrant_client.http import models as qmodels
+    from agent.tools.doc_search import _get_embedding_model
+
+    try:
+        model = _get_embedding_model()
+        output = model.encode(
+            [query],
+            batch_size=1,
+            max_length=512,
+            return_dense=True,
+            return_sparse=True,
+            return_colbert_vecs=False,
+        )
+        dense_vector = output["dense_vecs"][0].tolist()
+
+        from agent.tools.doc_search import _get_qdrant_client
+        client = _get_qdrant_client()
+
+        conf_filter = qmodels.Filter(
+            must=[
+                qmodels.FieldCondition(key="source_type", match=qmodels.MatchValue(value="confluence")),
+                qmodels.FieldCondition(key="team_id",     match=qmodels.MatchValue(value=team_id)),
+            ]
+        )
+
+        response = await client.query_points(
+            collection_name=settings.qdrant_collection,
+            query=dense_vector,
+            using=settings.qdrant_dense_vector_name,
+            query_filter=conf_filter,
+            limit=10,
+            with_payload=True,
+        )
+
+        chunks = []
+        for hit in response.points:
+            p = hit.payload or {}
+            chunks.append(RetrievedChunk(
+                chunk_id=p.get("chunk_id", str(hit.id)),
+                text=p.get("text", ""),
+                source=p.get("source", ""),
+                source_type="confluence",
+                score=hit.score,
+                metadata=p.get("metadata", {}),
+            ))
+
+        logger.info("confluence_search: found %d chunks for query=%r team=%s", len(chunks), query, team_id)
+        return chunks
+
+    except Exception:
+        logger.exception("confluence_search: search failed")
+        return []

agent/tools/doc_search.py

@@ -16,6 +16,16 @@ from agent.models import RetrievedChunk
 
 logger = logging.getLogger(__name__)
 
+
+def _get_qdrant_client() -> AsyncQdrantClient:
+    """Return Qdrant client — cloud if QDRANT_URL is set, otherwise local."""
+    if settings.qdrant_url:
+        return AsyncQdrantClient(
+            url=settings.qdrant_url,
+            api_key=settings.qdrant_api_key or None,
+        )
+    return AsyncQdrantClient(host=settings.qdrant_host, port=settings.qdrant_port)
+
 _PII_ENTITY_TYPES = [
     "person",
     "email",
@@ -33,6 +43,7 @@ _gliner = None
 _bm25_index: Optional[BM25Okapi] = None
 _bm25_corpus: Optional[list[str]] = None
 _bm25_doc_ids: Optional[list[str]] = None
+_bm25_metadata: Optional[list[dict]] = None
 
 
 def _get_embedding_model():
@@ -47,9 +58,17 @@ def _get_embedding_model():
 def _get_reranker():
     global _reranker
     if _reranker is None:
-        from FlagEmbedding import FlagReranker
-        logger.info("Loading BGE reranker: %s", settings.bge_reranker_model)
-        _reranker = FlagReranker(settings.bge_reranker_model, use_fp16=True)
+        import numpy as np
+        from sentence_transformers import CrossEncoder
+        logger.info("Loading CrossEncoder reranker (ms-marco-MiniLM-L-6-v2)")
+        _model = CrossEncoder('cross-encoder/ms-marco-MiniLM-L-6-v2')
+        class _Reranker:
+            def compute_score(self, pairs, normalize=True):
+                raw = _model.predict(pairs)
+                if normalize:
+                    return list(1 / (1 + np.exp(-raw)))
+                return list(raw.tolist())
+        _reranker = _Reranker()
     return _reranker
 
 
@@ -62,15 +81,15 @@ def _get_gliner():
     return _gliner
 
 
-def _load_bm25() -> tuple[Optional[BM25Okapi], Optional[list[str]], Optional[list[str]]]:
-    global _bm25_index, _bm25_corpus, _bm25_doc_ids
+def _load_bm25() -> tuple[Optional[BM25Okapi], Optional[list[str]], Optional[list[str]], Optional[list[dict]]]:
+    global _bm25_index, _bm25_corpus, _bm25_doc_ids, _bm25_metadata
     if _bm25_index is not None:
-        return _bm25_index, _bm25_corpus, _bm25_doc_ids
+        return _bm25_index, _bm25_corpus, _bm25_doc_ids, _bm25_metadata
 
     index_path = Path(settings.bm25_index_path)
     if not index_path.exists():
         logger.warning("BM25 index not found at %s — skipping BM25 retrieval", index_path)
-        return None, None, None
+        return None, None, None, None
 
     try:
         with index_path.open("rb") as f:
@@ -78,12 +97,13 @@ def _load_bm25() -> tuple[Optional[BM25Okapi], Optional[list[str]], Optional[lis
         _bm25_index = data["index"]
         _bm25_corpus = data["corpus"]
         _bm25_doc_ids = data["doc_ids"]
+        _bm25_metadata = data.get("metadata")
         logger.info("Loaded BM25 index with %d documents", len(_bm25_doc_ids))
     except Exception:
         logger.exception("Failed to load BM25 index from %s", index_path)
-        return None, None, None
+        return None, None, None, None
 
-    return _bm25_index, _bm25_corpus, _bm25_doc_ids
+    return _bm25_index, _bm25_corpus, _bm25_doc_ids, _bm25_metadata
 
 
 def _mask_pii(text: str) -> str:
@@ -133,7 +153,7 @@ async def run_doc_search(
     sparse_values = [sparse_weights[i] for i in sparse_indices]
 
     bm25_ranked_ids: list[str] = []
-    bm25, corpus, doc_ids = _load_bm25()
+    bm25, corpus, doc_ids, bm25_metadata = _load_bm25()
     if bm25 is not None and doc_ids:
         tokenized = masked_query.lower().split()
         bm25_scores = bm25.get_scores(tokenized)
@@ -147,7 +167,7 @@ async def run_doc_search(
     qdrant_score_map: dict[str, float] = {}
 
     try:
-        client = AsyncQdrantClient(host=settings.qdrant_host, port=settings.qdrant_port)
+        client = _get_qdrant_client()
 
         # Channel-based filter (RBAC): use allowed_channel_ids when set.
         # Falls back to team_id for legacy points that pre-date channel tagging.
@@ -224,7 +244,13 @@ async def run_doc_search(
             if doc_ids and doc_id in doc_ids:
                 idx = doc_ids.index(doc_id)
                 text = corpus[idx] if corpus else ""
-                payload = {"chunk_id": doc_id, "text": text, "source": "bm25", "source_type": "internal"}
+                meta = bm25_metadata[idx] if bm25_metadata else {}
+                payload = {
+                    "chunk_id": doc_id,
+                    "text": text,
+                    "source": meta.get("source") or meta.get("source_url") or "bm25",
+                    "source_type": meta.get("source_type", "internal"),
+                }
             else:
                 continue
         candidates.append({"id": doc_id, "payload": payload, "rrf_score": rrf_scores[doc_id]})

agent/tools/slack_search.py

@@ -0,0 +1,87 @@
+"""Slack search — searches conversation history using bot token (search.messages requires user token)."""
+
+from __future__ import annotations
+
+import logging
+import os
+
+import httpx
+
+from agent.models import RetrievedChunk
+
+logger = logging.getLogger(__name__)
+
+
+async def run_slack_search(query: str, team_id: str) -> list[RetrievedChunk]:
+    """Search Slack message history for the query by scanning channels the bot is in."""
+    token = os.environ.get("SLACK_BOT_TOKEN", "")
+    if not token:
+        logger.warning("slack_search: SLACK_BOT_TOKEN not set")
+        return []
+
+    keywords = [w.lower() for w in query.split() if len(w) > 3]
+    if not keywords:
+        keywords = query.lower().split()
+
+    try:
+        async with httpx.AsyncClient(timeout=15) as h:
+            # Get channels the bot is a member of
+            r = await h.get(
+                "https://slack.com/api/conversations.list",
+                headers={"Authorization": f"Bearer {token}"},
+                params={"types": "public_channel,private_channel", "limit": 200, "exclude_archived": True},
+            )
+        data = r.json()
+        if not data.get("ok"):
+            logger.warning("slack_search: conversations.list error: %s", data.get("error"))
+            return []
+
+        channels = [c for c in data.get("channels", []) if c.get("is_member")]
+        if not channels:
+            logger.info("slack_search: bot is not a member of any channels")
+            return []
+
+        chunks: list[RetrievedChunk] = []
+
+        for ch in channels[:10]:  # cap at 10 channels
+            ch_id   = ch["id"]
+            ch_name = ch.get("name", ch_id)
+            try:
+                async with httpx.AsyncClient(timeout=15) as h:
+                    r = await h.get(
+                        "https://slack.com/api/conversations.history",
+                        headers={"Authorization": f"Bearer {token}"},
+                        params={"channel": ch_id, "limit": 200},
+                    )
+                msgs = r.json().get("messages", [])
+            except Exception:
+                continue
+
+            for m in msgs:
+                text = m.get("text", "")
+                if not text:
+                    continue
+                text_lower = text.lower()
+                if any(kw in text_lower for kw in keywords):
+                    ts   = m.get("ts", "")
+                    user = m.get("user", m.get("username", "unknown"))
+                    link = f"https://slack.com/archives/{ch_id}/p{ts.replace('.', '')}"
+                    chunks.append(RetrievedChunk(
+                        chunk_id=f"slack-{ch_id}-{ts}",
+                        text=f"[#{ch_name}] {text}",
+                        source=link,
+                        source_type="slack",
+                        score=1.0,
+                        metadata={"channel": ch_name, "channel_id": ch_id, "user": user, "ts": ts},
+                    ))
+                    if len(chunks) >= 10:
+                        break
+            if len(chunks) >= 10:
+                break
+
+        logger.info("slack_search: found %d messages for query=%r", len(chunks), query)
+        return chunks
+
+    except Exception:
+        logger.exception("slack_search: search failed")
+        return []

agent/tools/sql_query.py

@@ -94,8 +94,8 @@ async def run_sql_query(
     team_id: str,
     allowed_channel_ids: list[str] | None = None,
 ) -> list[RetrievedChunk]:
-    if not settings.database_url:
-        logger.warning("sql_query: DATABASE_URL not configured — skipping")
+    if not settings.effective_database_url:
+        logger.warning("sql_query: DATABASE_URL / PG_DSN not configured — skipping")
         return []
 
     # --- Step 1: NL → SQL via Gemini Flash ---
@@ -128,7 +128,7 @@ async def run_sql_query(
         return []
 
     # Replace placeholder with positional parameter for asyncpg
-    parameterized_sql = raw_sql.replace("'<TEAM_ID_PLACEHOLDER>'", "$1")
+    parameterized_sql = re.sub(r"'<TEAM_ID_PLACEHOLDER>'", "$1", raw_sql, count=1)
 
     # --- Step 3: Safety validation ---
     ok, reason = _validate_sql(parameterized_sql)
@@ -140,7 +140,7 @@ async def run_sql_query(
     try:
         import asyncpg  # imported lazily — only needed when tool is active
 
-        conn = await asyncpg.connect(settings.database_url)
+        conn = await asyncpg.connect(settings.effective_database_url)
         try:
             rows = await conn.fetch(parameterized_sql, team_id)
         finally:

agent/tools/ticket_lookup.py

@@ -1,4 +1,4 @@
-"""Jira ticket lookup tool — interface stub, no live credentials yet."""
+"""Jira ticket lookup — searches ingested Jira chunks from Qdrant."""
 
 from __future__ import annotations
 
@@ -11,10 +11,57 @@ logger = logging.getLogger(__name__)
 
 
 async def run_ticket_lookup(query: str, team_id: str) -> list[RetrievedChunk]:
-    """Stub — returns empty until JIRA_BASE_URL and JIRA_API_TOKEN are configured."""
-    if not settings.jira_base_url or not settings.jira_api_token:
-        logger.info("ticket_lookup: Jira credentials not configured, returning empty results")
-        return []
+    """Search Qdrant for Jira chunks matching the query."""
+    from qdrant_client import AsyncQdrantClient
+    from qdrant_client.http import models as qmodels
+    from agent.tools.doc_search import _get_embedding_model
+
+    try:
+        model = _get_embedding_model()
+        output = model.encode(
+            [query],
+            batch_size=1,
+            max_length=512,
+            return_dense=True,
+            return_sparse=True,
+            return_colbert_vecs=False,
+        )
+        dense_vector = output["dense_vecs"][0].tolist()
+
+        from agent.tools.doc_search import _get_qdrant_client
+        client = _get_qdrant_client()
+
+        jira_filter = qmodels.Filter(
+            must=[
+                qmodels.FieldCondition(key="source_type", match=qmodels.MatchValue(value="jira")),
+                qmodels.FieldCondition(key="team_id",     match=qmodels.MatchValue(value=team_id)),
+            ]
+        )
 
-    logger.warning("ticket_lookup: stub returning empty results")
-    return []
+        response = await client.query_points(
+            collection_name=settings.qdrant_collection,
+            query=dense_vector,
+            using=settings.qdrant_dense_vector_name,
+            query_filter=jira_filter,
+            limit=10,
+            with_payload=True,
+        )
+
+        chunks = []
+        for hit in response.points:
+            p = hit.payload or {}
+            chunks.append(RetrievedChunk(
+                chunk_id=p.get("chunk_id", str(hit.id)),
+                text=p.get("text", ""),
+                source=p.get("source", ""),
+                source_type="jira",
+                score=hit.score,
+                metadata=p.get("metadata", {}),
+            ))
+
+        logger.info("ticket_lookup: found %d chunks for query=%r team=%s", len(chunks), query, team_id)
+        return chunks
+
+    except Exception:
+        logger.exception("ticket_lookup: search failed")
+        return []

frontend/src/App.tsx

@@ -47,7 +47,7 @@ export default function App() {
     const handler = (e: KeyboardEvent) => {
       if ((e.metaKey || e.ctrlKey) && e.key === 'k') {
         e.preventDefault()
-        navigate({ to: '/query', search: { q: undefined } })
+        navigate({ to: '/query', search: { q: undefined, qid: undefined, fresh: false } })
       }
     }
     window.addEventListener('keydown', handler)

frontend/src/components/admin/AdminDashboard.tsx

@@ -5,13 +5,15 @@ import { GraphIngestButton } from './GraphIngestButton'
 import { FileUploadWidget } from './FileUploadWidget'
 import { DataSourceManager } from './DataSourceManager'
 import { SystemLogs } from './SystemLogs'
+import { AdminUserManagement } from './AdminUserManagement'
 
-type Tab = 'overview' | 'data-sources' | 'ingest' | 'logs'
+type Tab = 'overview' | 'data-sources' | 'ingest' | 'users' | 'logs'
 
 const TABS: { id: Tab; label: string }[] = [
   { id: 'overview',      label: 'System Status' },
   { id: 'data-sources',  label: 'Data Sources'  },
   { id: 'ingest',        label: 'Ingest'         },
+  { id: 'users',         label: 'Users'          },
   { id: 'logs',          label: 'System Logs'    },
 ]
 
@@ -45,8 +47,8 @@ export function AdminDashboard() {
           <div className="rounded-xl border border-surface-subtle p-5">
             <p className="mb-4 text-sm font-medium text-stone-500">Sync Controls</p>
             <div className="flex flex-col gap-3">
-              <SyncTrigger source="jira"       sourceKey="ENG"  label="Jira (ENG)"         />
-              <SyncTrigger source="confluence" sourceKey="TEAM" label="Confluence (TEAM)"   />
+              <SyncTrigger source="jira"       sourceKey="KAN"      label="Jira (KAN)"           />
+              <SyncTrigger source="confluence" sourceKey="Godspeed" label="Confluence (Godspeed)" />
             </div>
           </div>
 
@@ -69,6 +71,12 @@ export function AdminDashboard() {
         </div>
       )}
 
+      {tab === 'users' && (
+        <div className="rounded-xl border border-surface-subtle">
+          <AdminUserManagement />
+        </div>
+      )}
+
       {tab === 'logs' && <SystemLogs />}
     </div>
   )

frontend/src/components/admin/AdminUserManagement.tsx

@@ -1,194 +1,181 @@
-import { useState, useRef } from 'react'
-import { useQuery, useMutation } from '@tanstack/react-query'
+import { useState } from 'react'
+import { useQuery, useMutation, useQueryClient } from '@tanstack/react-query'
 import { apiFetch } from '@/lib/http'
-import { UserInvite, BulkUserImport } from '@/types/settings'
-import { User } from '@/types/user'
+import { UserInvite } from '@/types/settings'
 import { LoadingSkeleton } from '../common/LoadingSkeleton'
 
+interface WorkspaceUser {
+  id: string
+  email: string
+  name: string
+  role: string
+  is_active: boolean
+}
+
 export function AdminUserManagement() {
+  const qc = useQueryClient()
   const [showAddForm, setShowAddForm] = useState(false)
-  const [showBulkImport, setShowBulkImport] = useState(false)
-  const [formData, setFormData] = useState<UserInvite>({
-    email: '',
-    name: '',
-    role: 'engineer',
-  })
-  const fileInputRef = useRef<HTMLInputElement>(null)
+  const [formData, setFormData] = useState<UserInvite>({ email: '', name: '', role: 'engineer' })
+  const [inviteLink, setInviteLink] = useState<{ url: string; emailSent: boolean } | null>(null)
+  const [error, setError] = useState<string | null>(null)
 
-  // Fetch users
-  const { data: users, isLoading, refetch } = useQuery({
+  const { data, isLoading } = useQuery({
     queryKey: ['workspace-users'],
     queryFn: async () => {
-      // TODO: GET /api/workspace/users
-      return [] as User[]
+      const res = await apiFetch('/api/workspace/users')
+      const json = await res.json()
+      return (json.users ?? []) as WorkspaceUser[]
     },
   })
+  const users = data ?? []
 
-  // Add single user
-  const { mutate: addUser, isPending: isAddingUser } = useMutation({
+  const { mutate: sendInvite, isPending: isSending } = useMutation({
     mutationFn: async (input: UserInvite) => {
-      // TODO: POST /api/workspace/users (send invitation)
-      console.log('Add user:', input)
+      const res = await apiFetch('/api/auth/invite', {
+        method: 'POST',
+        headers: { 'Content-Type': 'application/json' },
+        body: JSON.stringify({ email: input.email, name: input.name, role: input.role }),
+      })
+      if (!res.ok) {
+        const body = await res.json().catch(() => ({}))
+        throw new Error(body.detail ?? `Error ${res.status}`)
+      }
+      return res.json() as Promise<{ ok: boolean; email: string; invite_url: string; email_sent: boolean }>
     },
-    onSuccess: () => {
+    onSuccess: (data) => {
       setFormData({ email: '', name: '', role: 'engineer' })
-      setShowAddForm(false)
-      refetch()
+      setError(null)
+      setInviteLink({ url: data.invite_url, emailSent: data.email_sent })
+      qc.invalidateQueries({ queryKey: ['workspace-users'] })
+    },
+    onError: (err: Error) => {
+      setError(err.message)
     },
   })
 
-  // Bulk import users
-  const { mutate: bulkImport, isPending: isImporting } = useMutation({
-    mutationFn: async (file: File) => {
-      // TODO: POST /api/workspace/users/bulk-import with CSV/Excel parsing
-      console.log('Bulk import file:', file)
-      // For now, mock parsing
-      return { count: 0 } as { count: number }
-    },
-    onSuccess: () => {
-      setShowBulkImport(false)
-      if (fileInputRef.current) fileInputRef.current.value = ''
-      refetch()
+  const { mutate: removeUser } = useMutation({
+    mutationFn: async (userId: string) => {
+      const res = await apiFetch(`/api/workspace/users/${userId}`, { method: 'DELETE' })
+      if (!res.ok) throw new Error(`Error ${res.status}`)
     },
+    onSuccess: () => qc.invalidateQueries({ queryKey: ['workspace-users'] }),
   })
 
-  const handleAddUser = () => {
+  const handleInvite = () => {
     if (!formData.email.trim() || !formData.name.trim()) return
-    addUser(formData)
-  }
-
-  const handleBulkImport = async (e: React.ChangeEvent<HTMLInputElement>) => {
-    const file = e.currentTarget.files?.[0]
-    if (file) {
-      bulkImport(file)
-    }
+    setError(null)
+    setInviteLink(null as null)
+    sendInvite(formData)
   }
 
   return (
     <div className="space-y-6 p-6">
-      {/* Header with Actions */}
+      {/* Header */}
       <div className="flex items-center justify-between">
         <div>
           <h3 className="text-sm font-semibold">Workspace Users</h3>
-          <p className="mt-1 text-xs text-stone-500">{users?.length || 0} users total</p>
-        </div>
-        <div className="flex gap-2">
-          <button
-            onClick={() => setShowAddForm(!showAddForm)}
-            className="rounded bg-brand px-3 py-1.5 text-sm font-medium text-white hover:bg-brand-dark"
-          >
-            + Add User
-          </button>
-          <button
-            onClick={() => setShowBulkImport(!showBulkImport)}
-            className="rounded border border-stone-300 px-3 py-1.5 text-sm font-medium text-stone-700 hover:bg-stone-50 dark:border-stone-600 dark:text-stone-300 dark:hover:bg-stone-800"
-          >
-            📥 Import CSV
-          </button>
+          <p className="mt-1 text-xs text-stone-500">{users.length} user{users.length !== 1 ? 's' : ''}</p>
         </div>
+        <button
+          onClick={() => { setShowAddForm(!showAddForm); setInviteLink(null as null); setError(null) }}
+          className="rounded bg-brand px-3 py-1.5 text-sm font-medium text-white hover:bg-brand-dark"
+        >
+          + Invite User
+        </button>
       </div>
 
-      {/* Add User Form */}
+      {/* Invite Form */}
       {showAddForm && (
         <div className="space-y-4 rounded-lg border border-blue-200 bg-blue-50 p-4 dark:border-blue-900 dark:bg-stone-900">
-          <div>
-            <label className="block text-sm font-medium text-stone-700 dark:text-stone-300">
-              Name
-            </label>
-            <input
-              type="text"
-              value={formData.name}
-              onChange={(e) => setFormData({ ...formData, name: e.target.value })}
-              placeholder="John Doe"
-              className="mt-1 block w-full rounded border border-stone-300 bg-white px-3 py-2 text-stone-900 placeholder-stone-400 focus:border-brand focus:outline-none focus:ring-1 focus:ring-brand dark:border-stone-600 dark:bg-stone-800 dark:text-white"
-            />
-          </div>
-
-          <div>
-            <label className="block text-sm font-medium text-stone-700 dark:text-stone-300">
-              Email
-            </label>
-            <input
-              type="email"
-              value={formData.email}
-              onChange={(e) => setFormData({ ...formData, email: e.target.value })}
-              placeholder="john@example.com"
-              className="mt-1 block w-full rounded border border-stone-300 bg-white px-3 py-2 text-stone-900 placeholder-stone-400 focus:border-brand focus:outline-none focus:ring-1 focus:ring-brand dark:border-stone-600 dark:bg-stone-800 dark:text-white"
-            />
+          <div className="grid grid-cols-1 gap-3 sm:grid-cols-3">
+            <div>
+              <label className="block text-xs font-medium text-stone-700 dark:text-stone-300">Name</label>
+              <input
+                type="text"
+                value={formData.name}
+                onChange={(e) => setFormData({ ...formData, name: e.target.value })}
+                placeholder="Jane Smith"
+                className="mt-1 block w-full rounded border border-stone-300 bg-white px-3 py-2 text-sm text-stone-900 placeholder-stone-400 focus:border-brand focus:outline-none focus:ring-1 focus:ring-brand dark:border-stone-600 dark:bg-stone-800 dark:text-white"
+              />
+            </div>
+            <div>
+              <label className="block text-xs font-medium text-stone-700 dark:text-stone-300">Email</label>
+              <input
+                type="email"
+                value={formData.email}
+                onChange={(e) => setFormData({ ...formData, email: e.target.value })}
+                placeholder="jane@company.com"
+                className="mt-1 block w-full rounded border border-stone-300 bg-white px-3 py-2 text-sm text-stone-900 placeholder-stone-400 focus:border-brand focus:outline-none focus:ring-1 focus:ring-brand dark:border-stone-600 dark:bg-stone-800 dark:text-white"
+              />
+            </div>
+            <div>
+              <label className="block text-xs font-medium text-stone-700 dark:text-stone-300">Role</label>
+              <select
+                value={formData.role}
+                onChange={(e) => setFormData({ ...formData, role: e.target.value as UserInvite['role'] })}
+                className="mt-1 block w-full rounded border border-stone-300 bg-white px-3 py-2 text-sm text-stone-900 focus:border-brand focus:outline-none focus:ring-1 focus:ring-brand dark:border-stone-600 dark:bg-stone-800 dark:text-white"
+              >
+                <option value="engineer">Engineer</option>
+                <option value="manager">Manager</option>
+                <option value="admin">Admin</option>
+                <option value="org_admin">Org Admin</option>
+              </select>
+            </div>
           </div>
 
-          <div>
-            <label className="block text-sm font-medium text-stone-700 dark:text-stone-300">
-              Role
-            </label>
-            <select
-              value={formData.role}
-              onChange={(e) =>
-                setFormData({
-                  ...formData,
-                  role: e.target.value as 'engineer' | 'manager' | 'admin',
-                })
-              }
-              className="mt-1 block w-full rounded border border-stone-300 bg-white px-3 py-2 text-stone-900 focus:border-brand focus:outline-none focus:ring-1 focus:ring-brand dark:border-stone-600 dark:bg-stone-800 dark:text-white"
-            >
-              <option value="engineer">Engineer</option>
-              <option value="manager">Manager</option>
-              <option value="admin">Admin</option>
-            </select>
-          </div>
+          {error && (
+            <p className="rounded bg-red-100 px-3 py-2 text-xs text-red-700 dark:bg-red-900/30 dark:text-red-400">
+              {error}
+            </p>
+          )}
 
           <div className="flex gap-2">
             <button
-              onClick={handleAddUser}
-              disabled={isAddingUser || !formData.email.trim() || !formData.name.trim()}
+              onClick={handleInvite}
+              disabled={isSending || !formData.email.trim() || !formData.name.trim()}
               className="rounded bg-brand px-3 py-1.5 text-sm font-medium text-white hover:bg-brand-dark disabled:opacity-50"
             >
-              {isAddingUser ? 'Adding...' : 'Add User'}
+              {isSending ? 'Sending...' : 'Send Invite'}
             </button>
             <button
-              onClick={() => setShowAddForm(false)}
+              onClick={() => { setShowAddForm(false); setInviteLink(null as null); setError(null) }}
               className="rounded border border-stone-300 px-3 py-1.5 text-sm font-medium text-stone-700 hover:bg-stone-100 dark:border-stone-600 dark:text-stone-300 dark:hover:bg-stone-700"
             >
               Cancel
             </button>
           </div>
-        </div>
-      )}
 
-      {/* Bulk Import Form */}
-      {showBulkImport && (
-        <div className="space-y-4 rounded-lg border border-green-200 bg-green-50 p-4 dark:border-green-900 dark:bg-stone-900">
-          <div>
-            <label className="block text-sm font-medium text-stone-700 dark:text-stone-300">
-              Upload CSV or Excel file
-            </label>
-            <p className="mt-1 text-xs text-stone-600 dark:text-stone-400">
-              Format: name, email, role (engineer|manager|admin)
-            </p>
-            <input
-              ref={fileInputRef}
-              type="file"
-              accept=".csv,.xlsx,.xls"
-              onChange={handleBulkImport}
-              disabled={isImporting}
-              className="mt-2 block w-full text-sm text-stone-500 file:rounded file:border-0 file:bg-brand file:px-3 file:py-1.5 file:text-sm file:font-medium file:text-white hover:file:bg-brand-dark dark:text-stone-400"
-            />
-          </div>
-          <button
-            onClick={() => setShowBulkImport(false)}
-            className="rounded border border-stone-300 px-3 py-1.5 text-sm font-medium text-stone-700 hover:bg-stone-100 dark:border-stone-600 dark:text-stone-300 dark:hover:bg-stone-700"
-          >
-            Cancel
-          </button>
+          {inviteLink && (
+            <div className="rounded-lg border border-green-200 bg-green-50 p-3 dark:border-green-800 dark:bg-green-950/30">
+              <p className="mb-1 text-xs font-medium text-green-800 dark:text-green-300">
+                {inviteLink.emailSent
+                  ? 'Invite email sent! Share the link below as backup:'
+                  : 'Email not configured — share this link directly with the user:'}
+              </p>
+              <div className="flex items-center gap-2">
+                <input
+                  readOnly
+                  value={inviteLink.url}
+                  className="flex-1 rounded border border-green-300 bg-white px-2 py-1.5 text-xs font-mono text-stone-800 dark:border-green-700 dark:bg-stone-800 dark:text-stone-200"
+                />
+                <button
+                  onClick={() => navigator.clipboard.writeText(inviteLink.url)}
+                  className="rounded border border-green-300 px-2 py-1.5 text-xs font-medium text-green-700 hover:bg-green-100 dark:border-green-700 dark:text-green-400 dark:hover:bg-green-900/40"
+                >
+                  Copy
+                </button>
+              </div>
+              <p className="mt-1 text-xs text-green-700 dark:text-green-400">Expires in 7 days.</p>
+            </div>
+          )}
         </div>
       )}
 
-      {/* Users List */}
+      {/* Users Table */}
       <div>
-        <h4 className="mb-3 text-sm font-semibold">Active Users</h4>
         {isLoading ? (
           <LoadingSkeleton rows={5} className="h-10" />
-        ) : users && users.length > 0 ? (
+        ) : users.length > 0 ? (
           <div className="overflow-x-auto rounded-lg border border-stone-200 dark:border-stone-700">
             <table className="w-full text-left text-sm">
               <thead className="border-b border-stone-200 bg-stone-50 dark:border-stone-700 dark:bg-stone-800">
@@ -196,10 +183,8 @@ export function AdminUserManagement() {
                   <th className="px-4 py-2 font-semibold text-stone-700 dark:text-stone-300">Name</th>
                   <th className="px-4 py-2 font-semibold text-stone-700 dark:text-stone-300">Email</th>
                   <th className="px-4 py-2 font-semibold text-stone-700 dark:text-stone-300">Role</th>
-                  <th className="px-4 py-2 font-semibold text-stone-700 dark:text-stone-300">Team</th>
-                  <th className="px-4 py-2 text-right font-semibold text-stone-700 dark:text-stone-300">
-                    Actions
-                  </th>
+                  <th className="px-4 py-2 font-semibold text-stone-700 dark:text-stone-300">Status</th>
+                  <th className="px-4 py-2 text-right font-semibold text-stone-700 dark:text-stone-300">Actions</th>
                 </tr>
               </thead>
               <tbody className="divide-y divide-stone-200 dark:divide-stone-700">
@@ -212,9 +197,22 @@ export function AdminUserManagement() {
                         {u.role}
                       </span>
                     </td>
-                    <td className="px-4 py-3 text-stone-600 dark:text-stone-400">{u.team?.name || '—'}</td>
+                    <td className="px-4 py-3">
+                      <span className={`rounded-full px-2 py-0.5 text-xs font-medium ${
+                        u.is_active
+                          ? 'bg-green-100 text-green-700 dark:bg-green-900/30 dark:text-green-400'
+                          : 'bg-stone-100 text-stone-500 dark:bg-stone-700 dark:text-stone-400'
+                      }`}>
+                        {u.is_active ? 'Active' : 'Inactive'}
+                      </span>
+                    </td>
                     <td className="px-4 py-3 text-right">
-                      <button className="text-xs font-medium text-stone-500 hover:text-red-600">
+                      <button
+                        onClick={() => {
+                          if (confirm(`Remove ${u.name} from workspace?`)) removeUser(u.id)
+                        }}
+                        className="text-xs font-medium text-stone-500 hover:text-red-600"
+                      >
                         Remove
                       </button>
                     </td>
@@ -226,7 +224,7 @@ export function AdminUserManagement() {
         ) : (
           <div className="rounded-lg border-2 border-dashed border-stone-300 p-6 text-center dark:border-stone-600">
             <p className="text-sm text-stone-500">No users yet</p>
-            <p className="mt-1 text-xs text-stone-400">Add your first user to get started</p>
+            <p className="mt-1 text-xs text-stone-400">Click "Invite User" to add your first team member</p>
           </div>
         )}
       </div>

frontend/src/components/admin/SyncTrigger.tsx

@@ -1,23 +1,70 @@
-import { useState } from 'react'
+import { useState, useEffect, useRef } from 'react'
 import { apiFetch, ApiError } from '@/lib/http'
 import { useUIStore } from '@/stores/uiStore'
 import { cn } from '@/lib/utils'
 
+type JobStatus = 'pending' | 'running' | 'completed' | 'failed'
+
 interface Props {
   source:    'jira' | 'confluence'
-  /** project key for Jira, space key for Confluence */
   sourceKey: string
   label:     string
   lastSynced?: string
 }
 
+const STATUS_LABEL: Record<JobStatus, string> = {
+  pending:   'Queued…',
+  running:   'Syncing…',
+  completed: 'Sync complete',
+  failed:    'Sync failed',
+}
+
+const STATUS_COLOR: Record<JobStatus, string> = {
+  pending:   'text-stone-400',
+  running:   'text-blue-500',
+  completed: 'text-green-600',
+  failed:    'text-red-500',
+}
+
 export function SyncTrigger({ source, sourceKey, label, lastSynced }: Props) {
-  const [taskId, setTaskId]   = useState<string | null>(null)
-  const [loading, setLoading] = useState(false)
-  const addToast              = useUIStore((s) => s.addToast)
+  const [taskId,    setTaskId]    = useState<string | null>(null)
+  const [jobStatus, setJobStatus] = useState<JobStatus | null>(null)
+  const [loading,   setLoading]   = useState(false)
+  const addToast = useUIStore((s) => s.addToast)
+  const pollRef  = useRef<ReturnType<typeof setInterval> | null>(null)
+
+  // Poll job status until terminal state
+  useEffect(() => {
+    if (!taskId) return
+    if (pollRef.current) clearInterval(pollRef.current)
+
+    const poll = async () => {
+      try {
+        const res  = await apiFetch(`/ingest/jobs/${taskId}`)
+        const data = await res.json() as { status: JobStatus }
+        setJobStatus(data.status)
+        if (data.status === 'completed' || data.status === 'failed') {
+          clearInterval(pollRef.current!)
+          pollRef.current = null
+          addToast({
+            type:    data.status === 'completed' ? 'success' : 'error',
+            message: data.status === 'completed'
+              ? `${label} sync finished`
+              : `${label} sync failed`,
+          })
+        }
+      } catch { /* ignore poll errors */ }
+    }
+
+    poll() // immediate first check
+    pollRef.current = setInterval(poll, 4000)
+    return () => { if (pollRef.current) clearInterval(pollRef.current) }
+  }, [taskId]) // eslint-disable-line react-hooks/exhaustive-deps
 
   const trigger = async () => {
     setLoading(true)
+    setTaskId(null)
+    setJobStatus(null)
     try {
       const path = source === 'jira'
         ? `/jira/sync/${sourceKey}`
@@ -25,39 +72,53 @@ export function SyncTrigger({ source, sourceKey, label, lastSynced }: Props) {
       const res  = await apiFetch(path, { method: 'POST' })
       const data = await res.json() as { task_id: string }
       setTaskId(data.task_id)
+      setJobStatus('pending')
       addToast({ type: 'success', message: `${label} sync queued` })
     } catch (err) {
       if (!(err instanceof ApiError)) {
         addToast({ type: 'error', message: `Failed to trigger ${label} sync` })
       }
-      // ApiError toasts are handled in apiFetch
     } finally {
       setLoading(false)
     }
   }
 
+  const isBusy = loading || jobStatus === 'pending' || jobStatus === 'running'
+
   return (
     <div className="flex items-center justify-between rounded-lg border border-surface-subtle p-4">
       <div>
         <p className="text-sm font-medium">{label}</p>
-        {lastSynced && (
+        {lastSynced && !jobStatus && (
           <p className="text-xs text-stone-500">Last synced {lastSynced}</p>
         )}
+        {jobStatus && (
+          <div className="mt-1 flex items-center gap-1.5">
+            {(jobStatus === 'pending' || jobStatus === 'running') && (
+              <span className="inline-block h-2.5 w-2.5 animate-spin rounded-full border-2 border-blue-300 border-t-blue-600" />
+            )}
+            <p className={cn('text-xs font-medium', STATUS_COLOR[jobStatus])}>
+              {STATUS_LABEL[jobStatus]}
+            </p>
+          </div>
+        )}
         {taskId && (
-          <p className="mt-0.5 font-mono text-xs text-stone-400">Task {taskId.slice(0, 8)}…</p>
+          <p className="mt-0.5 font-mono text-[10px] text-stone-400">
+            Task {taskId.slice(0, 8)}…
+          </p>
         )}
       </div>
       <button
         onClick={trigger}
-        disabled={loading}
+        disabled={isBusy}
         className={cn(
           'rounded-lg px-3 py-1.5 text-xs font-medium transition-colors',
-          loading
+          isBusy
             ? 'cursor-not-allowed bg-stone-100 text-stone-400 dark:bg-stone-800'
             : 'bg-brand text-white hover:bg-brand-dark',
         )}
       >
-        {loading ? 'Queuing…' : 'Sync now'}
+        {loading ? 'Queuing…' : isBusy ? 'Syncing…' : 'Sync now'}
       </button>
     </div>
   )

frontend/src/components/common/Sidebar.tsx

@@ -165,7 +165,7 @@ export function Sidebar() {
               {recent.map((item) => (
                 <button
                   key={item.id}
-                  onClick={() => navigate({ to: '/query', search: { q: item.query } })}
+                  onClick={() => navigate({ to: '/query', search: { q: item.query, qid: undefined, fresh: false } })}
                   title={item.query}
                   className="flex w-full items-center gap-2 rounded-lg px-2.5 py-1.5 text-left transition-colors hover:bg-stone-50 dark:hover:bg-stone-900"
                 >
@@ -175,7 +175,7 @@ export function Sidebar() {
                       item.success ? 'bg-green-400' : 'bg-stone-300',
                     )}
                   />
-                  <span className="truncate text-xs text-stone-500 dark:text-stone-400">
+                  <span className="truncate text-xs text-stone-500 dark:text-stone-400" title={item.query}>
                     {item.query.length > 34 ? item.query.slice(0, 34) + '…' : item.query}
                   </span>
                 </button>

frontend/src/components/query/SuggestedTopics.tsx

@@ -3,11 +3,11 @@ interface Props {
 }
 
 const DEFAULTS = [
-  'What services does the auth team own?',
-  'Show me recent incidents affecting the payments service',
-  'Which libraries are deprecated in our stack?',
-  'What is the onboarding process for new engineers?',
-  'Summarise open Jira tickets for the infra team',
+  'What is Godspeed and what does it do?',
+  'How do I set up Godspeed locally?',
+  'What are the main components of the Godspeed architecture?',
+  'What API endpoints does Godspeed expose?',
+  'What is the incident runbook for Godspeed?',
 ]
 
 export function SuggestedTopics({ onSelect }: Props) {

frontend/src/components/results/KnowledgeGraph.tsx

@@ -68,6 +68,13 @@ export function KnowledgeGraph({ nodes, edges, streaming, onNodeClick, onNodeHov
       if (!containerRef.current) return
 
       const el = containerRef.current
+
+      // Track real mouse position since force-graph's onNodeHover doesn't provide it
+      let mouseX = 0
+      let mouseY = 0
+      const trackMouse = (e: MouseEvent) => { mouseX = e.clientX; mouseY = e.clientY }
+      el.addEventListener('mousemove', trackMouse)
+
       const fg = ForceGraph2D()
       fg(el)
       fg.backgroundColor('transparent')
@@ -85,7 +92,7 @@ export function KnowledgeGraph({ nodes, edges, streaming, onNodeClick, onNodeHov
           onNodeClick({ id: n.id, label: n.label, name: n.name })
         })
         .onNodeHover((n: FGNode | null) => {
-          onNodeHover(n ? { id: n.id, label: n.label, name: n.name } : null, 0, 0)
+          onNodeHover(n ? { id: n.id, label: n.label, name: n.name } : null, mouseX, mouseY)
         })
       graphRef.current = fg
 
@@ -108,6 +115,7 @@ export function KnowledgeGraph({ nodes, edges, streaming, onNodeClick, onNodeHov
       ro?.disconnect()
       graphRef.current?._destructor?.()
       graphRef.current = null
+      // trackMouse listener is on el which is removed from DOM — GC handles it
     }
     // eslint-disable-next-line react-hooks/exhaustive-deps
   }, [])

frontend/src/components/results/ResultsPage.tsx

@@ -1,5 +1,5 @@
 import { useCallback, useEffect, useRef, useState } from 'react'
-import { useSearch } from '@tanstack/react-router'
+import { useSearch, useNavigate } from '@tanstack/react-router'
 import { useAuthStore } from '@/stores/authStore'
 import { useUIStore } from '@/stores/uiStore'
 import { useSSEStream } from '@/hooks/useSSEStream'
@@ -27,13 +27,28 @@ import type {
 
 type Tab = 'graph' | 'answer'
 
+const QR_PREFIX = 'gs_qr_'
+
+interface Exchange {
+  id: string
+  query: string
+  answer: string
+  citations: RetrievedChunk[]
+  plan: AgentTask[]
+}
+
 export function ResultsPage() {
   const user       = useAuthStore((s) => s.user)
+  const navigate   = useNavigate()
   const { graphCollapsed, toggleGraphCollapsed } = useUIStore()
   const sessionRef = useRef(crypto.randomUUID())
-  const { q: initialQuery } = useSearch({ from: '/query' })
+  const savedRef   = useRef(false)
+  const { q: initialQuery, qid: initialQid, fresh: forceRun } = useSearch({ from: '/query' })
 
-  // ── SSE state ───────────────────────────────────────────────────────────────
+  // ── Conversation history ─────────────────────────────────────────────────────
+  const [exchanges, setExchanges] = useState<Exchange[]>([])
+
+  // ── SSE state (current streaming exchange) ───────────────────────────────────
   const [plan, setPlan]               = useState<AgentTask[]>([])
   const [agentStatuses, setStatuses]  = useState<Record<string, AgentStatus>>({})
   const [answerText, setAnswerText]   = useState('')
@@ -43,28 +58,39 @@ export function ResultsPage() {
   const [queryId, setQueryId]           = useState('')
   const [shareOpen, setShareOpen]       = useState(false)
 
+  // Refs to read current values inside the completion effect without stale closure
+  const answerTextRef  = useRef('')
+  const citationsRef   = useRef<RetrievedChunk[]>([])
+  const planRef        = useRef<AgentTask[]>([])
+  const currentQueryRef = useRef('')
+
+  // Keep refs in sync
+  useEffect(() => { answerTextRef.current  = answerText  }, [answerText])
+  useEffect(() => { citationsRef.current   = citations   }, [citations])
+  useEffect(() => { planRef.current        = plan        }, [plan])
+  useEffect(() => { currentQueryRef.current = currentQuery }, [currentQuery])
+
   // ── Graph state ─────────────────────────────────────────────────────────────
   const [graphNodes, setGraphNodes]     = useState<GraphNode[]>([])
   const [graphEdges, setGraphEdges]     = useState<GraphEdge[]>([])
   const [hoveredNode, setHoveredNode]   = useState<GraphNode | null>(null)
   const [hoverPos, setHoverPos]         = useState({ x: 0, y: 0 })
   const [selectedNode, setSelectedNode] = useState<GraphNode | null>(null)
-  // Accumulated in refs to avoid setState on every 50ms node message
   const nodesAccRef = useRef<GraphNode[]>([])
   const edgesAccRef = useRef<GraphEdge[]>([])
 
   // ── Mobile tab ──────────────────────────────────────────────────────────────
   const [activeTab, setActiveTab]       = useState<Tab>('graph')
-  // Local — not persisted; maximize is a session-level power-user action
   const [graphMaximized, setGraphMaximized] = useState(false)
 
   // ── Hooks ───────────────────────────────────────────────────────────────────
   const { state, error, firstEventArrived, stream } = useSSEStream()
   const { gState, retryCount, connect, disconnect } = useGraphStream()
 
-  // ── Shared graph reconnect ───────────────────────────────────────────────────
-  // Used both by runQuery (new query) and the standalone "Reload / Try again" buttons.
-  // Clears accumulated state then opens a fresh WS to /graph/stream.
+  useEffect(() => {
+    if (gState === 'error' && !graphCollapsed) toggleGraphCollapsed()
+  }, [gState]) // eslint-disable-line react-hooks/exhaustive-deps
+
   const reconnectGraph = useCallback(() => {
     nodesAccRef.current = []
     edgesAccRef.current = []
@@ -88,7 +114,21 @@ export function ResultsPage() {
   // ── Query runner ─────────────────────────────────────────────────────────────
   const runQuery = useCallback(
     async (query: string) => {
-      // Reset all state for the new query
+      // Save the current completed exchange to history before resetting
+      if (answerTextRef.current && currentQueryRef.current) {
+        setExchanges((prev) => [
+          ...prev,
+          {
+            id:        crypto.randomUUID(),
+            query:     currentQueryRef.current,
+            answer:    answerTextRef.current,
+            citations: citationsRef.current,
+            plan:      planRef.current,
+          },
+        ])
+      }
+
+      savedRef.current = false
       setPlan([])
       setStatuses({})
       setAnswerText('')
@@ -97,7 +137,7 @@ export function ResultsPage() {
       setCurrentQuery(query)
       setQueryId(crypto.randomUUID())
       setShareOpen(false)
-      setActiveTab('graph') // always start on graph tab so user sees it populate
+      setActiveTab('graph')
       setSelectedNode(null)
 
       reconnectGraph()
@@ -153,13 +193,65 @@ export function ResultsPage() {
     setSelectedNode(node)
   }, [])
 
-  // Auto-fire from URL ?q= on mount
   const runQueryRef = useRef(runQuery)
   runQueryRef.current = runQuery
+  const textKey = (q: string) => QR_PREFIX + 'q:' + q
+
+  const restoreFromCache = (raw: string): boolean => {
+    try {
+      const { query, answer, citations: cc } = JSON.parse(raw)
+      setCurrentQuery(query)
+      setAnswerText(answer)
+      setCitations(cc ?? [])
+      savedRef.current = true
+      return true
+    } catch { return false }
+  }
+
+  const lastParamKeyRef = useRef('')
+
   useEffect(() => {
+    const paramKey = `${initialQid ?? ''}|${initialQuery ?? ''}|${String(forceRun)}`
+    if (lastParamKeyRef.current === paramKey) return
+    lastParamKeyRef.current = paramKey
+
+    if (!forceRun) {
+      if (initialQid) {
+        try {
+          const raw = sessionStorage.getItem(QR_PREFIX + initialQid)
+          if (raw && restoreFromCache(raw)) return
+        } catch { /* ignore */ }
+      }
+      if (initialQuery) {
+        try {
+          const raw = sessionStorage.getItem(textKey(initialQuery))
+          if (raw && restoreFromCache(raw)) return
+        } catch { /* ignore */ }
+      }
+    }
     if (initialQuery) runQueryRef.current(initialQuery)
     // eslint-disable-next-line react-hooks/exhaustive-deps
-  }, [])
+  }, [initialQuery, initialQid, forceRun])
+
+  // Cache completed result and update URL to ?qid=
+  useEffect(() => {
+    if (state !== 'complete' || !currentQuery || !answerText || savedRef.current) return
+    savedRef.current = true
+    const qid = sessionRef.current
+    const payload = JSON.stringify({ query: currentQuery, answer: answerText, citations })
+    try {
+      sessionStorage.setItem(QR_PREFIX + qid, payload)
+      sessionStorage.setItem(textKey(currentQuery), payload)
+    } catch { /* storage full */ }
+    navigate({ to: '/query', search: { qid, q: undefined, fresh: false }, replace: true })
+  // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [state])
+
+  // Scroll to bottom when new exchange added or answer streams in
+  const bottomRef = useRef<HTMLDivElement>(null)
+  useEffect(() => {
+    bottomRef.current?.scrollIntoView({ behavior: 'smooth', block: 'nearest' })
+  }, [exchanges.length, answerText])
 
   // ── Derived ─────────────────────────────────────────────────────────────────
   const isLoading   = state === 'loading'
@@ -167,36 +259,34 @@ export function ResultsPage() {
   const isComplete  = state === 'complete'
   const isError     = state === 'error'
   const isActive    = isLoading || isStreaming
-  const hasData     = firstEventArrived.current  // set when first SSE event arrives
+  const hasData     = firstEventArrived.current
 
-  // ── Tab indicators ──────────────────────────────────────────────────────────
   const graphHasContent  = graphNodes.length > 0
   const answerHasContent = !!answerText || plan.length > 0
 
   return (
     <div className="mx-auto flex min-h-screen max-w-7xl flex-col gap-6 px-4 py-8">
 
-      {/* Search bar — always visible */}
+      {/* Search bar */}
       <SearchBox
+        key={currentQuery || initialQuery}
         onSubmit={runQuery}
         disabled={isActive}
-        defaultValue={initialQuery ?? ''}
+        defaultValue={currentQuery || (initialQuery ?? '')}
       />
 
-      {/* Idle home state — no query yet */}
-      {state === 'idle' && !currentQuery && (
+      {/* Idle home state */}
+      {state === 'idle' && !currentQuery && exchanges.length === 0 && (
         <div className="flex flex-1 items-center justify-center">
           <p className="text-sm text-stone-400">Ask a question to get started.</p>
         </div>
       )}
 
-      {/* ── Results layout — mounted immediately when query fires ─────────── */}
-      {currentQuery && (
+      {/* ── Results layout ────────────────────��────────────────────────────── */}
+      {(currentQuery || exchanges.length > 0) && (
         <>
-          {/* Mobile tab bar — hidden on lg where both columns are visible */}
-          {/* Answer tab is disabled (grey) until first SSE event arrives */}
+          {/* Mobile tab bar */}
           <div className="flex rounded-xl border border-surface-subtle bg-stone-50 p-1 dark:bg-stone-800/50 lg:hidden">
-            {/* Graph tab — always accessible; shows animation even before nodes arrive */}
             <button
               onClick={() => setActiveTab('graph')}
               className={cn(
@@ -213,14 +303,12 @@ export function ResultsPage() {
                 </span>
               )}
             </button>
-
-            {/* Answer tab — disabled (grey, not clickable) until first SSE event */}
             <button
               onClick={() => setActiveTab('answer')}
-              disabled={!hasData}
+              disabled={!hasData && exchanges.length === 0}
               className={cn(
                 'flex flex-1 items-center justify-center gap-2 rounded-lg py-2 text-sm font-medium transition-colors',
-                !hasData
+                !hasData && exchanges.length === 0
                   ? 'cursor-not-allowed text-stone-300 dark:text-stone-600'
                   : activeTab === 'answer'
                   ? 'bg-white text-stone-900 shadow-sm dark:bg-stone-700 dark:text-stone-100'
@@ -228,26 +316,23 @@ export function ResultsPage() {
               )}
             >
               Answer
-              {/* Pulsing dot only once answer has data and is still streaming */}
               {hasData && isActive && answerHasContent && (
                 <span className="h-1.5 w-1.5 animate-pulse rounded-full bg-brand" />
               )}
             </button>
           </div>
 
-          {/* Two-column grid — collapses to single when graph is hidden */}
+          {/* Two-column grid */}
           <div className={cn(
             'flex flex-col gap-6 lg:grid lg:items-start lg:gap-8',
             graphCollapsed ? 'lg:grid-cols-1' : 'lg:grid-cols-[1fr_380px]',
           )}>
 
-            {/* ── Left: answer column ──────────────────────────────────────── */}
+            {/* ── Left: conversation column ─────────────────────────────────── */}
             <div className={cn(
-              'flex flex-col gap-6',
+              'flex flex-col gap-8',
               activeTab === 'graph' && 'hidden lg:flex',
             )}>
-
-              {/* Show-graph chip — visible on desktop when graph panel is collapsed */}
               {graphCollapsed && (
                 <button
                   onClick={toggleGraphCollapsed}
@@ -257,88 +342,98 @@ export function ResultsPage() {
                 </button>
               )}
 
-              {/* Thinking indicator — before any SSE event arrives */}
-              {isActive && !hasData && (
-                <div className="flex items-center gap-2 text-sm text-stone-400">
-                  <span className="inline-block h-4 w-4 animate-spin rounded-full border-2 border-stone-200 border-t-stone-500" />
-                  Thinking…
+              {/* ── Completed past exchanges ─────────────────────────────── */}
+              {exchanges.map((ex) => (
+                <div key={ex.id} className="flex flex-col gap-4 border-b border-surface-subtle pb-8">
+                  <p className="text-sm font-semibold text-stone-500">{ex.query}</p>
+                  <Answer text={ex.answer} />
+                  {ex.citations.length > 0 && <Citations chunks={ex.citations} />}
                 </div>
-              )}
-
-              {/* Agent progress badges */}
-              {plan.length > 0 && (
-                <AgentBadges plan={plan} statuses={agentStatuses} />
-              )}
-
-              {/* Streaming answer */}
-              {answerText && <Answer text={answerText} />}
-
-              {/* Complete with no answer */}
-              {isComplete && !answerText && <NoResultsState query={currentQuery} />}
-
-              {/* Guardrail escalation */}
-              {guardrail?.escalate && <HallucinationWarning />}
-
-              {/* Hard error — no answer at all */}
-              {isError && !answerText && (
-                <TimeoutError
-                  message={error ?? undefined}
-                  onRetry={() => runQuery(currentQuery)}
-                />
-              )}
-
-              {/* Soft error — answer cut short mid-stream */}
-              {isError && answerText && error?.includes('incomplete') && (
-                <div className="flex items-center gap-2 rounded-lg border border-amber-200 bg-amber-50 px-4 py-3 text-sm text-amber-700 dark:border-amber-800 dark:bg-amber-950/30 dark:text-amber-300">
-                  <span aria-hidden>⚠</span>
-                  Answer may be incomplete — connection dropped mid-stream.
-                  <button
-                    onClick={() => runQuery(currentQuery)}
-                    className="ml-auto shrink-0 underline"
-                  >
-                    Retry
-                  </button>
+              ))}
+
+              {/* ── Current streaming exchange ────────────────────────────── */}
+              {currentQuery && (
+                <div className="flex flex-col gap-4">
+                  {/* Show query heading only when there are prior exchanges */}
+                  {exchanges.length > 0 && (
+                    <p className="text-sm font-semibold text-stone-700 dark:text-stone-300">
+                      {currentQuery}
+                    </p>
+                  )}
+
+                  {isActive && !hasData && (
+                    <div className="flex items-center gap-2 text-sm text-stone-400">
+                      <span className="inline-block h-4 w-4 animate-spin rounded-full border-2 border-stone-200 border-t-stone-500" />
+                      Thinking…
+                    </div>
+                  )}
+
+                  {plan.length > 0 && (
+                    <AgentBadges plan={plan} statuses={agentStatuses} />
+                  )}
+
+                  {answerText && <Answer text={answerText} />}
+
+                  {isComplete && !answerText && citations.length === 0 && (
+                    <NoResultsState query={currentQuery} />
+                  )}
+
+                  {guardrail?.escalate && <HallucinationWarning />}
+
+                  {isError && !answerText && (
+                    <TimeoutError
+                      message={error ?? undefined}
+                      onRetry={() => runQuery(currentQuery)}
+                    />
+                  )}
+
+                  {isError && answerText && error?.includes('incomplete') && (
+                    <div className="flex items-center gap-2 rounded-lg border border-amber-200 bg-amber-50 px-4 py-3 text-sm text-amber-700 dark:border-amber-800 dark:bg-amber-950/30 dark:text-amber-300">
+                      <span aria-hidden>⚠</span>
+                      Answer may be incomplete — connection dropped mid-stream.
+                      <button onClick={() => runQuery(currentQuery)} className="ml-auto shrink-0 underline">
+                        Retry
+                      </button>
+                    </div>
+                  )}
+
+                  {citations.length > 0 && <Citations chunks={citations} />}
+
+                  {isComplete && citations.length > 3 && (
+                    <RelatedDocs chunks={citations.slice(3)} />
+                  )}
+
+                  {isComplete && (
+                    <div className="flex items-center justify-between gap-4">
+                      <QueryFeedback queryId={queryId} />
+                      <button
+                        onClick={() => setShareOpen(true)}
+                        className="shrink-0 text-xs text-stone-400 underline hover:text-stone-600"
+                      >
+                        Share
+                      </button>
+                    </div>
+                  )}
                 </div>
               )}
 
-              {/* Citations */}
-              {citations.length > 0 && <Citations chunks={citations} />}
+              <div ref={bottomRef} />
 
-              {/* Related docs (overflow citations) */}
-              {isComplete && citations.length > 3 && (
-                <RelatedDocs chunks={citations.slice(3)} />
-              )}
-
-              {/* Footer actions */}
-              {isComplete && (
-                <div className="flex items-center justify-between gap-4">
-                  <QueryFeedback queryId={queryId} />
-                  <button
-                    onClick={() => setShareOpen(true)}
-                    className="shrink-0 text-xs text-stone-400 underline hover:text-stone-600"
-                  >
-                    Share
-                  </button>
-                </div>
+              {/* Follow-up always visible at bottom once there's any content */}
+              {(isComplete || exchanges.length > 0) && (
+                <FollowUp onSubmit={runQuery} disabled={isActive} />
               )}
-
-              {isComplete && <FollowUp onSubmit={runQuery} />}
             </div>
 
             {/* ── Right: graph panel ───────────────────────────────────────── */}
-            {/* On mobile: controlled by activeTab. On desktop: hidden when graphCollapsed */}
             <div className={cn(
               'flex flex-col overflow-hidden rounded-xl border border-surface-subtle',
-              // Mobile: toggle by tab; desktop: toggle by collapsed state
               activeTab === 'answer' ? 'hidden lg:flex' : 'flex',
               graphCollapsed && 'lg:hidden',
-              // Maximized: fixed fullscreen overlay
               graphMaximized
                 ? 'fixed inset-0 z-50 rounded-none border-0 bg-white dark:bg-stone-950'
                 : 'h-[440px]',
             )}>
-
-              {/* Graph toolbar */}
               <div className="flex shrink-0 items-center gap-2 border-b border-surface-subtle px-3 py-2">
                 <span className="flex-1 text-xs font-semibold uppercase tracking-wide text-stone-400">
                   Knowledge Graph
@@ -348,7 +443,6 @@ export function ResultsPage() {
                     {graphNodes.length} nodes · {graphEdges.length} edges
                   </span>
                 )}
-                {/* Maximize / restore */}
                 <button
                   onClick={() => setGraphMaximized((m) => !m)}
                   title={graphMaximized ? 'Exit fullscreen' : 'Fullscreen'}
@@ -357,7 +451,6 @@ export function ResultsPage() {
                 >
                   {graphMaximized ? '⤡' : '⤢'}
                 </button>
-                {/* Collapse — desktop only; on mobile graph visibility is via tabs */}
                 <button
                   onClick={toggleGraphCollapsed}
                   title="Collapse graph"
@@ -368,7 +461,6 @@ export function ResultsPage() {
                 </button>
               </div>
 
-              {/* Canvas — flex-1 fills remaining panel height */}
               <KnowledgeGraph
                 nodes={graphNodes}
                 edges={graphEdges}
@@ -378,7 +470,6 @@ export function ResultsPage() {
                 className="flex-1"
               />
 
-              {/* Graph footer */}
               <div className="flex shrink-0 items-center gap-2 border-t border-surface-subtle px-3 py-2">
                 {gState === 'done' && (
                   <button
@@ -392,7 +483,6 @@ export function ResultsPage() {
                 {gState === 'retrying' && <NetworkRetry attempt={retryCount + 1} />}
               </div>
 
-              {/* Max-retries error bar */}
               {gState === 'error' && (
                 <div className="flex shrink-0 items-center justify-between border-t border-stone-200 bg-stone-50 px-3 py-2 text-sm dark:border-stone-700 dark:bg-stone-800/40">
                   <span className="text-stone-500 dark:text-stone-400">
@@ -411,17 +501,14 @@ export function ResultsPage() {
         </>
       )}
 
-      {/* Tooltip — document-level fixed positioning */}
       <GraphNodeTooltip node={hoveredNode} x={hoverPos.x} y={hoverPos.y} />
 
-      {/* Share modal */}
       <ShareResults
         query={currentQuery}
         open={shareOpen}
         onClose={() => setShareOpen(false)}
       />
 
-      {/* Node detail slide-in panel */}
       <GraphNodeDetailPanel
         node={selectedNode}
         teamId={user?.team_id ?? 'default'}

frontend/src/components/workspace/QueryHistory.tsx

@@ -1,4 +1,4 @@
-import { useState } from 'react'
+import { useState, useEffect, useRef } from 'react'
 import { useQuery } from '@tanstack/react-query'
 import { apiFetch } from '@/lib/http'
 import { LoadingSkeleton } from '@/components/common/LoadingSkeleton'
@@ -25,11 +25,13 @@ async function fetchHistory(page: number): Promise<HistoryResponse> {
 
 interface Props {
   onReplay?: (query: string) => void
+  focusId?:  string
 }
 
-export function QueryHistory({ onReplay }: Props) {
+export function QueryHistory({ onReplay, focusId }: Props) {
   const [page,     setPage]     = useState(1)
-  const [expanded, setExpanded] = useState<string | null>(null)
+  const [expanded, setExpanded] = useState<string | null>(focusId ?? null)
+  const focusRef = useRef<HTMLDivElement>(null)
 
   const { data, isLoading } = useQuery({
     queryKey:  ['workspace-history', page],
@@ -37,6 +39,13 @@ export function QueryHistory({ onReplay }: Props) {
     staleTime: 60_000,
   })
 
+  // Scroll the focused item into view once data loads
+  useEffect(() => {
+    if (focusId && focusRef.current) {
+      focusRef.current.scrollIntoView({ behavior: 'smooth', block: 'center' })
+    }
+  }, [focusId, data])
+
   const totalPages = data ? Math.ceil(data.total / 20) : 1
 
   return (
@@ -61,7 +70,11 @@ export function QueryHistory({ onReplay }: Props) {
           {(data?.items ?? []).map((item) => (
             <div
               key={item.id}
-              className="rounded-xl border border-surface-subtle transition-colors hover:border-stone-300 dark:hover:border-stone-600"
+              ref={item.id === focusId ? focusRef : undefined}
+              className={cn(
+                'rounded-xl border border-surface-subtle transition-colors hover:border-stone-300 dark:hover:border-stone-600',
+                item.id === focusId && 'border-brand/50 dark:border-brand/40',
+              )}
             >
               <button
                 className="flex w-full items-start gap-3 p-4 text-left"

frontend/src/config/env.ts

@@ -1,11 +1,14 @@
-const get = (key: string): string => {
-  // eslint-disable-next-line @typescript-eslint/no-explicit-any
-  const val = (import.meta as any).env[key]
-  if (!val) throw new Error(`Missing env var: ${key}`)
-  return val
-}
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+const _env = (import.meta as any).env
 
 export const env = {
-  apiBaseUrl: get('VITE_API_BASE_URL'),
-  wsBaseUrl:  get('VITE_WS_BASE_URL'),
-} as const
+  // Empty string → relative URLs → works when frontend is served from FastAPI via ngrok
+  apiBaseUrl: _env.VITE_API_BASE_URL || '',
+
+  // Derived at runtime from page host so WebSocket works on any ngrok URL automatically
+  get wsBaseUrl(): string {
+    if (_env.VITE_WS_BASE_URL) return _env.VITE_WS_BASE_URL
+    const proto = window.location.protocol === 'https:' ? 'wss:' : 'ws:'
+    return `${proto}//${window.location.host}`
+  },
+}

frontend/src/config/routes.ts

@@ -65,7 +65,9 @@ export const queryRoute = createRoute({
   beforeLoad:         requireAuth,
   component:          QueryPage,
   validateSearch: (s: Record<string, unknown>) => ({
-    q: typeof s.q === 'string' ? s.q : undefined,
+    q:     typeof s.q   === 'string' ? s.q   : undefined,
+    qid:   typeof s.qid === 'string' ? s.qid : undefined,
+    fresh: s.fresh === true || s.fresh === 'true',
   }),
 })
 
@@ -88,6 +90,9 @@ export const workspaceRoute = createRoute({
   path: '/workspace',
   beforeLoad:     requireAuth,
   component:      Workspace,
+  validateSearch: (s: Record<string, unknown>) => ({
+    id: typeof s.id === 'string' ? s.id : undefined,
+  }),
 })
 
 export const settingsRoute = createRoute({

frontend/src/hooks/useGraphStream.ts

@@ -2,7 +2,7 @@ import { useCallback, useEffect, useRef, useState } from 'react'
 import { env } from '@/config/env'
 import type { GraphNode, GraphEdge, GraphDoneEvent } from '@/types/api'
 
-const MAX_RETRIES   = 5
+const MAX_RETRIES   = 1
 const BASE_DELAY_MS = 1000
 
 type Callbacks = {
@@ -49,6 +49,11 @@ export function useGraphStream() {
           activeRef.current = false
           setGState('done')
           callbacksRef.current?.onDone(msg as GraphDoneEvent)
+        } else if (msg.event === 'error') {
+          // Server signalled a terminal error — stop retrying
+          activeRef.current = false
+          setGState('error')
+          callbacksRef.current?.onError(msg.message ?? 'Graph unavailable')
         }
       } catch {
         // Non-JSON frame — ignore

frontend/src/hooks/useNotifications.ts

@@ -21,9 +21,11 @@ export function useNotifications() {
 
     ws.onmessage = (evt) => {
       try {
-        const msg = JSON.parse(evt.data as string) as Omit<AppNotification, 'id' | 'read'>
+        const msg = JSON.parse(evt.data as string)
+        // Skip keepalive pings and any frame without a real message
+        if (!msg.message || msg.type === 'ping') return
         const notification: AppNotification = {
-          ...msg,
+          ...(msg as Omit<AppNotification, 'id' | 'read'>),
           id:   crypto.randomUUID(),
           read: false,
         }

frontend/src/pages/AcceptInvitePage.tsx

@@ -93,7 +93,7 @@ export default function AcceptInvitePage() {
             {...register('password')}
             type="password"
             autoComplete="new-password"
-            className="rounded border border-surface-subtle px-3 py-2 focus:outline-brand"
+            className="rounded border border-surface-subtle bg-white px-3 py-2 text-stone-900 placeholder-stone-400 focus:outline-brand dark:border-stone-600 dark:bg-stone-800 dark:text-white"
           />
           {errors.password && <span className="text-xs text-red-600">{errors.password.message}</span>}
         </label>
@@ -104,7 +104,7 @@ export default function AcceptInvitePage() {
             {...register('confirm')}
             type="password"
             autoComplete="new-password"
-            className="rounded border border-surface-subtle px-3 py-2 focus:outline-brand"
+            className="rounded border border-surface-subtle bg-white px-3 py-2 text-stone-900 placeholder-stone-400 focus:outline-brand dark:border-stone-600 dark:bg-stone-800 dark:text-white"
           />
           {errors.confirm && <span className="text-xs text-red-600">{errors.confirm.message}</span>}
         </label>

frontend/src/pages/Home.tsx

@@ -6,7 +6,7 @@ export default function Home() {
   const navigate = useNavigate()
 
   const handleQuery = (query: string) => {
-    navigate({ to: '/query', search: { q: query } })
+    navigate({ to: '/query', search: { q: query, qid: undefined, fresh: false } })
   }
 
   return (

frontend/src/pages/LoginPage.tsx

@@ -44,7 +44,7 @@ export default function LoginPage() {
             {...register('email')}
             type="email"
             autoComplete="email"
-            className="rounded border border-surface-subtle px-3 py-2 focus:outline-brand"
+            className="rounded border border-surface-subtle bg-white px-3 py-2 text-stone-900 placeholder-stone-400 focus:outline-brand dark:border-stone-600 dark:bg-stone-800 dark:text-white"
           />
           {errors.email && <span className="text-red-600 text-xs">{errors.email.message}</span>}
         </label>
@@ -55,7 +55,7 @@ export default function LoginPage() {
             {...register('password')}
             type="password"
             autoComplete="current-password"
-            className="rounded border border-surface-subtle px-3 py-2 focus:outline-brand"
+            className="rounded border border-surface-subtle bg-white px-3 py-2 text-stone-900 placeholder-stone-400 focus:outline-brand dark:border-stone-600 dark:bg-stone-800 dark:text-white"
           />
           {errors.password && <span className="text-red-600 text-xs">{errors.password.message}</span>}
         </label>

frontend/src/pages/OAuthCallbackPage.tsx

@@ -15,7 +15,11 @@ export default function OAuthCallbackPage() {
     const hasError = params.get('error') !== null
 
     if (hasError) {
-      addToast({ type: 'error', message: 'Google sign-in failed. Please try again.' })
+      const error = params.get('error')
+      const message = error === 'not_invited'
+        ? 'Your Google account is not linked to any workspace. Ask an admin to invite you first.'
+        : 'Google sign-in failed. Please try again.'
+      addToast({ type: 'error', message })
       navigate({ to: '/login' })
       return
     }

frontend/src/pages/WorkspacePage.tsx

@@ -1,17 +1,18 @@
-import { useNavigate } from '@tanstack/react-router'
+import { useNavigate, useSearch } from '@tanstack/react-router'
 import { QueryHistory } from '@/components/workspace/QueryHistory'
 
 export default function WorkspacePage() {
   const navigate = useNavigate()
+  const { id: focusId } = useSearch({ from: '/workspace' })
 
   const handleReplay = (query: string) => {
-    navigate({ to: '/query', search: { q: query } })
+    navigate({ to: '/query', search: { q: query, qid: undefined, fresh: true } })
   }
 
   return (
     <div className="mx-auto max-w-3xl px-4 py-8">
       <h1 className="mb-6 text-2xl font-semibold">Workspace</h1>
-      <QueryHistory onReplay={handleReplay} />
+      <QueryHistory onReplay={handleReplay} focusId={focusId} />
     </div>
   )
 }

frontend/src/types/api.ts

@@ -5,7 +5,7 @@ export interface QueryInput {
 }
 
 export interface AgentTask {
-  agent:      'doc_search' | 'ticket_lookup' | 'live_docs' | 'summariser'
+  agent:      'doc_search' | 'ticket_lookup' | 'confluence_search' | 'slack_search' | 'live_docs' | 'summariser' | 'sql_query'
   input:      string
   depends_on: string[]
 }

graph_store/stream.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import asyncio
 import logging
+import time
 
 from fastapi import APIRouter, WebSocket, WebSocketDisconnect
 from neo4j import AsyncGraphDatabase
@@ -11,6 +12,31 @@ from graph_store.config import settings
 logger = logging.getLogger(__name__)
 router = APIRouter(tags=["graph"])
 
+# After a DNS/connection failure, skip Neo4j for this many seconds before retrying
+_NEO4J_COOLDOWN_S = 120
+_neo4j_failed_at: float | None = None
+
+
+def _neo4j_is_available() -> bool:
+    global _neo4j_failed_at
+    if _neo4j_failed_at is None:
+        return True
+    if time.monotonic() - _neo4j_failed_at > _NEO4J_COOLDOWN_S:
+        _neo4j_failed_at = None
+        return True
+    return False
+
+
+def _mark_neo4j_failed() -> None:
+    global _neo4j_failed_at
+    if _neo4j_failed_at is None:
+        logger.warning(
+            "graph_stream: Neo4j unreachable (%s) — graph panel disabled for %ds",
+            settings.neo4j_uri,
+            _NEO4J_COOLDOWN_S,
+        )
+    _neo4j_failed_at = time.monotonic()
+
 # ---------------------------------------------------------------------------
 # Cypher queries
 # ---------------------------------------------------------------------------
@@ -126,15 +152,25 @@ async def graph_stream(websocket: WebSocket):
     else:
         logger.info("graph_stream: no channel restriction — serving full graph")
 
+    if not _neo4j_is_available():
+        try:
+            await websocket.send_json({"event": "error", "message": "Knowledge graph unavailable"})
+        except Exception:
+            pass
+        try:
+            await websocket.close()
+        except Exception:
+            pass
+        return
+
     driver = AsyncGraphDatabase.driver(
         settings.neo4j_uri,
         auth=(settings.neo4j_username, settings.neo4j_password),
         max_connection_lifetime=300,
-        connection_acquisition_timeout=60,
+        connection_acquisition_timeout=10,
         keep_alive=True,
     )
     try:
-        # Fetch all records at once (same pattern as traverse — avoids async-for issues)
         async with driver.session(database=settings.neo4j_database) as session:
             if use_filtered:
                 result = await session.run(
@@ -190,10 +226,10 @@ async def graph_stream(websocket: WebSocket):
 
     except WebSocketDisconnect:
         logger.info("graph_stream: client disconnected")
-    except Exception:
-        logger.exception("graph_stream: error")
+    except Exception as exc:
+        _mark_neo4j_failed()
         try:
-            await websocket.send_json({"event": "error", "message": "Graph stream failed"})
+            await websocket.send_json({"event": "error", "message": "Knowledge graph unavailable"})
         except Exception:
             pass
     finally:

ingestion/jobs/ingest_job.py

@@ -16,7 +16,11 @@ _SOURCE_REGISTRY: dict[str, Any] = {}
 
 
 async def _run_graph_pipeline(embedded, doc) -> None:
-    from graph_store.config import settings as graph_settings
+    try:
+        from graph_store.config import settings as graph_settings
+    except Exception:
+        logger.warning("ingest_job: graph_store not available — skipping graph pipeline")
+        return
 
     if not graph_settings.neo4j_uri:
         logger.warning("ingest_job: NEO4J_URI not set — skipping graph pipeline")
@@ -126,10 +130,10 @@ async def _run_ingest_async(job_id: str, payload: IngestSourcePayload) -> dict[s
             for chunk in chunks:
                 chunk.text = mask_pii(chunk.text)
 
+            upsert_document(doc, client=sb)
             embedded = embed_chunks(chunks)
             qdrant_upsert(embedded)
             sb_upsert_chunks(chunks, client=sb)
-            upsert_document(doc, client=sb)
             total_chunks += len(chunks)
             logger.info("ingest_job: ingested %d chunks for doc_id=%s", len(chunks), doc.doc_id)
 

ingestion/sources/github.py

@@ -37,7 +37,7 @@ class GithubSource(BaseSource):
         self._team_id = team_id
         self._repo_url = repo_url
         self._supabase = supabase_client
-        self._path_filter = path_filter or settings.github_path_filter
+        self._path_filter = path_filter  # empty string = no filter; caller sets explicitly
         self._branch = branch or settings.github_branch
         self._token = token or settings.github_token
         self._owner, self._repo = _parse_owner_repo(repo_url)
@@ -125,12 +125,15 @@ class GithubSource(BaseSource):
             logger.exception("github: failed to fetch file tree")
             return []
 
+        INDEXABLE_EXTENSIONS = {".md", ".txt", ".rst", ".py", ".ts", ".tsx", ".js", ".yaml", ".yml"}
+
         md_paths = [
             item["path"]
             for item in tree
             if item["type"] == "blob"
-            and item["path"].endswith(".md")
-            and item["path"].startswith(self._path_filter)
+            and any(item["path"].endswith(ext) for ext in INDEXABLE_EXTENSIONS)
+            and (not self._path_filter or item["path"].startswith(self._path_filter))
+            and item.get("size", 0) <= 200_000
         ]
 
         docs: list[RawDocument] = []
@@ -151,7 +154,7 @@ class GithubSource(BaseSource):
             source_url = data.get("html_url", f"{self._repo_url}/blob/{self._branch}/{path}")
             return RawDocument(
                 doc_id=doc_id,
-                title=path.split("/")[-1].removesuffix(".md"),
+                title=path.split("/")[-1].rsplit(".", 1)[0],
                 content=content,
                 source_url=source_url,
                 source_type="github",

ingestion/storage/bm25_store.py

@@ -24,15 +24,17 @@ class BM25Store:
             return []
         return re.findall(r"\b\w+\b", text.lower())
 
-    def rebuild_index(self, chunk_ids: List[str], texts: List[str]) -> None:
+    def rebuild_index(self, chunk_ids: List[str], texts: List[str], metadata: List[dict] | None = None) -> None:
         if not chunk_ids or not texts:
             logger.warning("BM25: empty corpus provided")
             return
         if len(chunk_ids) != len(texts):
             raise ValueError(f"chunk_ids ({len(chunk_ids)}) != texts ({len(texts)})")
+        if metadata is not None and len(metadata) != len(chunk_ids):
+            raise ValueError(f"metadata ({len(metadata)}) != chunk_ids ({len(chunk_ids)})")
 
-        cleaned_ids, cleaned_texts, tokenized_corpus = [], [], []
-        for chunk_id, text in zip(chunk_ids, texts):
+        cleaned_ids, cleaned_texts, tokenized_corpus, cleaned_metadata = [], [], [], []
+        for i, (chunk_id, text) in enumerate(zip(chunk_ids, texts)):
             if not chunk_id or not text or not text.strip():
                 continue
             tokens = self.tokenize(text)
@@ -41,12 +43,18 @@ class BM25Store:
             cleaned_ids.append(chunk_id)
             cleaned_texts.append(text)
             tokenized_corpus.append(tokens)
+            if metadata is not None:
+                cleaned_metadata.append(metadata[i])
 
         if not tokenized_corpus:
             raise ValueError("No valid documents to index")
 
+        payload = {"index": BM25Okapi(tokenized_corpus), "doc_ids": cleaned_ids, "corpus": cleaned_texts}
+        if metadata is not None:
+            payload["metadata"] = cleaned_metadata
+
         with self.index_path.open("wb") as f:
-            pickle.dump({"index": BM25Okapi(tokenized_corpus), "doc_ids": cleaned_ids, "corpus": cleaned_texts}, f)
+            pickle.dump(payload, f)
 
         logger.info("BM25 rebuilt with %d documents -> %s", len(cleaned_ids), self.index_path)
 
@@ -78,4 +86,5 @@ def rebuild_from_supabase() -> None:
     store.rebuild_index(
         chunk_ids=[r["chunk_id"] for r in rows],
         texts=[r["text"] for r in rows],
+        metadata=[{"source": r.get("source", ""), "source_type": r.get("source_type", "internal")} for r in rows],
     )

ingestion/storage/supabase_store.py

@@ -134,7 +134,7 @@ def update_cag_snapshot(team_id: str, snapshot: str, client: Optional[Client] =
 def get_all_chunks(client: Optional[Client] = None) -> list[dict[str, Any]]:
     sb = client or get_client()
     try:
-        result = sb.table("chunks").select("chunk_id, text").execute()
+        result = sb.table("chunks").select("chunk_id, text, source, source_type").execute()
         return result.data or []
     except Exception:
         logger.exception("supabase_store: failed to fetch all chunks")

main.py

@@ -77,7 +77,7 @@ app.include_router(tools_router)
 
 
 # ---------------------------------------------------------------------------
-# Health endpoint — pinged by admin dashboard + Docker healthcheck
+# Health endpoint — must be registered BEFORE the SPA catch-all
 # ---------------------------------------------------------------------------
 
 @app.get("/health", tags=["infra"])
@@ -130,3 +130,22 @@ async def health() -> dict:
         results["status"] = "degraded"
 
     return results
+
+
+# ---------------------------------------------------------------------------
+# Serve React build — SPA catch-all MUST be last (catches everything else)
+# ---------------------------------------------------------------------------
+import os as _os
+from fastapi.staticfiles import StaticFiles
+from fastapi.responses import FileResponse as _FileResponse
+
+_dist = _os.path.join(_os.path.dirname(__file__), "frontend", "dist")
+if _os.path.exists(_dist):
+    app.mount("/assets", StaticFiles(directory=_os.path.join(_dist, "assets")), name="assets")
+
+    @app.get("/{full_path:path}", include_in_schema=False)
+    async def serve_spa(full_path: str):
+        file = _os.path.join(_dist, full_path)
+        if _os.path.isfile(file):
+            return _FileResponse(file)
+        return _FileResponse(_os.path.join(_dist, "index.html"))

migrate_qdrant.py

@@ -0,0 +1,58 @@
+"""Migrate local Qdrant collection to Qdrant Cloud. Run once before deploying."""
+import asyncio, os, sys
+
+COLLECTION = "knowledge_base"
+CLOUD_URL = os.getenv("QDRANT_CLOUD_URL", "")
+CLOUD_KEY = os.getenv("QDRANT_CLOUD_API_KEY", "")
+
+if not CLOUD_URL or not CLOUD_KEY:
+    print("Set QDRANT_CLOUD_URL and QDRANT_CLOUD_API_KEY env vars first")
+    sys.exit(1)
+
+async def migrate():
+    from qdrant_client import AsyncQdrantClient
+    from qdrant_client.models import PointStruct
+
+    src = AsyncQdrantClient(host="localhost", port=6333)
+    dst = AsyncQdrantClient(url=CLOUD_URL, api_key=CLOUD_KEY)
+
+    info = await src.get_collection(COLLECTION)
+    print(f"Source collection: {info.points_count} points")
+
+    try:
+        await dst.delete_collection(COLLECTION)
+    except Exception:
+        pass
+
+    # Create with both dense and sparse vector configs
+    await dst.create_collection(
+        COLLECTION,
+        vectors_config=info.config.params.vectors,
+        sparse_vectors_config=info.config.params.sparse_vectors,
+    )
+    print("Created cloud collection (dense + sparse)")
+
+    offset = None
+    total = 0
+    while True:
+        records, offset = await src.scroll(
+            COLLECTION, offset=offset, limit=50,
+            with_payload=True, with_vectors=True
+        )
+        if not records:
+            break
+        structs = [
+            PointStruct(id=r.id, vector=r.vector, payload=r.payload or {})
+            for r in records
+            if r.vector is not None
+        ]
+        if structs:
+            await dst.upsert(COLLECTION, points=structs)
+        total += len(structs)
+        print(f"Migrated {total} points...")
+        if offset is None:
+            break
+
+    print(f"Done — {total} points migrated to {CLOUD_URL}")
+
+asyncio.run(migrate())

requirements.txt

@@ -17,10 +17,13 @@
 langgraph==0.2.76
 langchain-core==0.3.86
 langchain-google-genai==2.0.9
+langchain-openai==1.2.1
 
 
 # ── Embeddings & reranking ────────────────────────────────────────────────────
-FlagEmbedding==1.3.0
+FlagEmbedding==1.4.0
+# CrossEncoder reranker — replaces BGE reranker (Python 3.14 compatible)
+sentence-transformers==5.4.1
 
 
 # ── PII masking (local — zero egress) ────────────────────────────────────────
@@ -68,9 +71,10 @@ google-auth==2.37.0
 numpy==1.26.4
 # torch: CPU build. For GPU replace with the matching CUDA wheel:
 #   https://pytorch.org/get-started/locally/
-# torch 2.6 is compatible with numpy 1.26.x and transformers 4.51.x.
+# torch 2.6 is compatible with numpy 1.26.x and transformers 5.1.x.
+# transformers is constrained to <5.2.0 by gliner 0.2.x.
 torch==2.6.0
-transformers==4.51.0
+transformers==5.1.0
 
 
 # ── Document parsing ─────────────────────────────────────────────────────────
@@ -135,7 +139,7 @@ watchdog==6.0.0
 #   Ubuntu: apt install tesseract-ocr
 #   Windows: https://github.com/UB-Mannheim/tesseract/wiki
 pytesseract==0.3.13
-Pillow==11.1.0
+Pillow==12.2.0
 
 
 # ── Optional: live web fetching (configure via .env) ─────────────────────────
@@ -143,5 +147,8 @@ Pillow==11.1.0
 # firecrawl-py==1.9.0
 # tavily-python==0.5.0
 
+# ── Auth — password hashing ───────────────────────────────────────────────────
+bcrypt==5.0.0
+
 # ── NL-to-SQL agent — direct PostgreSQL connection ────────────────────────────
 asyncpg>=0.29.0

src/admin/router.py

@@ -31,24 +31,26 @@ async def _redis() -> aioredis.Redis:
 def _default_sources() -> list[dict]:
     """Build source list from env vars that are set."""
     sources = []
-    if os.getenv("JIRA_INSTANCE_URL") or settings.integrations.jira_instance_url:
+    jira_url = os.getenv("JIRA_BASE_URL") or os.getenv("JIRA_INSTANCE_URL") or settings.integrations.jira_instance_url
+    if jira_url:
         sources.append({
             "id":          "jira-default",
             "type":        "jira",
             "name":        "Jira",
-            "url":         settings.integrations.jira_instance_url or "https://your-org.atlassian.net",
-            "enabled":     bool(settings.integrations.jira_api_token),
+            "url":         jira_url,
+            "enabled":     bool(os.getenv("JIRA_API_TOKEN") or settings.integrations.jira_api_token),
             "last_sync":   None,
             "sync_status": "idle",
             "error_msg":   None,
         })
-    if os.getenv("CONFLUENCE_URL") or os.getenv("CONFLUENCE_SPACE_KEY"):
+    confluence_url = os.getenv("CONFLUENCE_BASE_URL") or os.getenv("CONFLUENCE_URL")
+    if confluence_url:
         sources.append({
             "id":          "confluence-default",
             "type":        "confluence",
             "name":        "Confluence",
-            "url":         os.getenv("CONFLUENCE_URL", "https://your-org.atlassian.net/wiki"),
-            "enabled":     bool(os.getenv("CONFLUENCE_API_TOKEN")),
+            "url":         confluence_url,
+            "enabled":     bool(os.getenv("CONFLUENCE_TOKEN") or os.getenv("CONFLUENCE_API_TOKEN")),
             "last_sync":   None,
             "sync_status": "idle",
             "error_msg":   None,

src/admin/users_api.py

@@ -52,7 +52,7 @@ async def list_users(user=Depends(require_role("admin", "org_admin"))) -> dict:
         result = (
             _client()
             .table("users")
-            .select("id, email, name, role, is_active, is_new_hire, workspace_id, team_id")
+            .select("id, email, name, role, is_active")
             .eq("workspace_id", DEFAULT_WORKSPACE_ID)
             .order("name")
             .execute()

src/auth/email.py

@@ -13,23 +13,26 @@ logger = logging.getLogger(__name__)
 
 def send_email(to: str, subject: str, html: str) -> None:
     """Fire-and-forget — logs on failure, never raises."""
-    if not settings.smtp_host or not settings.smtp_user:
-        logger.warning("email_not_configured — skipping send to %s", to)
-        return
+    threading.Thread(target=lambda: send_email_sync(to, subject, html), daemon=True).start()
 
-    def _send():
-        try:
-            msg = MIMEMultipart("alternative")
-            msg["Subject"] = subject
-            msg["From"] = settings.smtp_from
-            msg["To"] = to
-            msg.attach(MIMEText(html, "html"))
-            with smtplib.SMTP(settings.smtp_host, settings.smtp_port) as s:
-                s.starttls()
-                s.login(settings.smtp_user, settings.smtp_password)
-                s.sendmail(settings.smtp_from, to, msg.as_string())
-            logger.info("email_sent to=%s subject=%s", to, subject)
-        except Exception:
-            logger.exception("email_send_failed to=%s", to)
 
-    threading.Thread(target=_send, daemon=True).start()
+def send_email_sync(to: str, subject: str, html: str) -> bool:
+    """Synchronous send — returns True if sent, False if skipped or failed."""
+    if not settings.smtp_host or not settings.smtp_user:
+        logger.warning("email_not_configured — skipping send to %s", to)
+        return False
+    try:
+        msg = MIMEMultipart("alternative")
+        msg["Subject"] = subject
+        msg["From"]    = settings.smtp_from
+        msg["To"]      = to
+        msg.attach(MIMEText(html, "html"))
+        with smtplib.SMTP(settings.smtp_host, settings.smtp_port) as s:
+            s.starttls()
+            s.login(settings.smtp_user, settings.smtp_password)
+            s.sendmail(settings.smtp_from, to, msg.as_string())
+        logger.info("email_sent to=%s subject=%s", to, subject)
+        return True
+    except Exception:
+        logger.exception("email_send_failed to=%s", to)
+        return False

src/auth/router.py

@@ -432,10 +432,11 @@ async def send_invite(
         f"<p><a href=\"{invite_url}\">Accept your invitation</a></p>"
         f"<p>This link expires in 7 days.</p>"
     )
-    send_email(to=email, subject="You're invited to GodSpeed", html=html)
+    from src.auth.email import send_email_sync
+    email_sent = send_email_sync(to=email, subject="You're invited to GodSpeed", html=html)
 
-    logger.info("invite_sent to=%s by=%s role=%s", email, caller["id"], body.role)
-    return {"ok": True, "email": email}
+    logger.info("invite_sent to=%s by=%s role=%s email_sent=%s", email, caller["id"], body.role, email_sent)
+    return {"ok": True, "email": email, "invite_url": invite_url, "email_sent": email_sent}
 
 
 @router.get("/invite/{token}")

src/confluence_agent/tasks.py

@@ -15,7 +15,7 @@ def confluence_process_page(self, page_id: str, space_key: str = "", team_id: st
     """Webhook-triggered single-page ingestion."""
     try:
         from src.confluence_agent.pipeline import ingest_page
-        count = asyncio.get_event_loop().run_until_complete(
+        count = asyncio.run(
             ingest_page(page_id, space_key, team_id or confluence_config.team_id)
         )
         return {"page_id": page_id, "chunks_stored": count}
@@ -29,7 +29,7 @@ def confluence_sync_space(self, space_key: str, team_id: str = "") -> dict:
     """Full space sync."""
     try:
         from src.confluence_agent.pipeline import ingest_space
-        count = asyncio.get_event_loop().run_until_complete(
+        count = asyncio.run(
             ingest_space(space_key, team_id or confluence_config.team_id)
         )
         return {"space_key": space_key, "chunks_stored": count}
@@ -65,6 +65,6 @@ def confluence_periodic_sync() -> dict:
             results[space_key] = len(docs)
         return total
 
-    total = asyncio.get_event_loop().run_until_complete(_sync_all())
+    total = asyncio.run(_sync_all())
     logger.info("confluence_periodic_sync: synced %d pages, %d total chunks", sum(results.values()), total)
     return {"spaces": results, "total_chunks": total}

src/jira_agent/tasks.py

@@ -14,7 +14,7 @@ def jira_process_issue(self, issue_key: str, team_id: str = "") -> dict:
     """Webhook-triggered single-issue ingestion."""
     try:
         from src.jira_agent.pipeline import ingest_issue
-        count = asyncio.get_event_loop().run_until_complete(
+        count = asyncio.run(
             ingest_issue(issue_key, team_id or jira_config.team_id)
         )
         return {"issue_key": issue_key, "chunks_stored": count}
@@ -28,7 +28,7 @@ def jira_sync_project(self, project_key: str, team_id: str = "") -> dict:
     """Full project sync — intended for manual trigger or scheduled use."""
     try:
         from src.jira_agent.pipeline import ingest_project
-        count = asyncio.get_event_loop().run_until_complete(
+        count = asyncio.run(
             ingest_project(project_key, team_id or jira_config.team_id)
         )
         return {"project_key": project_key, "chunks_stored": count}

start-ngrok.sh

@@ -0,0 +1,50 @@
+#!/bin/bash
+# Start GodSpeed backend and ngrok tunnel
+
+# Kill any existing ngrok / uvicorn
+pkill -f "ngrok http" 2>/dev/null
+pkill -f "uvicorn" 2>/dev/null
+sleep 1
+
+echo "Starting ngrok..."
+ngrok http 8000 --log=stdout > /tmp/ngrok.log 2>&1 &
+NGROK_PID=$!
+
+# Wait for ngrok to be ready
+for i in $(seq 1 15); do
+  NGROK_URL=$(curl -s http://localhost:4040/api/tunnels 2>/dev/null | python3 -c "
+import sys, json
+try:
+    data = json.load(sys.stdin)
+    for t in data.get('tunnels', []):
+        if t.get('proto') == 'https':
+            print(t['public_url'])
+            break
+except: pass
+" 2>/dev/null)
+  [ -n "$NGROK_URL" ] && break
+  sleep 1
+done
+
+if [ -z "$NGROK_URL" ]; then
+  echo "ERROR: Could not get ngrok URL. Is ngrok authenticated? Run: ngrok config add-authtoken <token>"
+  kill $NGROK_PID 2>/dev/null
+  exit 1
+fi
+
+echo "Starting backend with FRONTEND_URL=$NGROK_URL ..."
+FRONTEND_URL=$NGROK_URL .venv/bin/uvicorn main:app --host 0.0.0.0 --port 8000 &
+BACKEND_PID=$!
+
+sleep 2
+
+echo ""
+echo "========================================="
+echo "  GodSpeed is live at: $NGROK_URL"
+echo "  Share this URL with your friends!"
+echo "========================================="
+echo ""
+echo "Press Ctrl+C to stop everything"
+
+trap "kill $BACKEND_PID $NGROK_PID 2>/dev/null; exit" INT
+wait