backend/memory/episodic_memory.py

# backend/memory/episodic_memory.py
import logging
import asyncio
from uuid import uuid4

try:
    import chromadb
    from chromadb.config import Settings
except ImportError:
    chromadb = None
    Settings = None

try:
    import google.generativeai as genai
    from config import GEMINI_API_KEY
    # Initialize genai explicitly
    genai.configure(api_key=GEMINI_API_KEY)
except ImportError:
    pass

# "models/embedding-001" was RETIRED — it 404s ("not found for API version
# v1beta"), which meant every embed() call fell through to the zero-vector
# fallback below. That failed silently and is worse than an outright error:
# identical all-zero vectors make every document equidistant, so semantic
# retrieval returns essentially arbitrary results while still "working".
# Verified live against this key — the only models exposing embedContent are
# gemini-embedding-001, gemini-embedding-2 and gemini-embedding-2-preview.
EMBED_MODEL = "models/gemini-embedding-001"
EMBED_DIM = 768  # keep in step with the existing persona_* collections


async def embed(text: str) -> list[float]:
    """Generates an embedding for the given text using Gemini."""
    try:
        def _get_embedding():
            result = genai.embed_content(
                model=EMBED_MODEL,
                content=text,
                task_type="retrieval_document",
                output_dimensionality=EMBED_DIM,
            )
            return result['embedding']

        vec = await asyncio.to_thread(_get_embedding)
        if not vec:
            raise ValueError("empty embedding returned")
        return vec
    except Exception as e:
        # Loud, not silent: a zero vector destroys ranking, so make it obvious
        # in the logs that retrieval quality is degraded rather than fine.
        logging.error(
            f"Embedding failed via {EMBED_MODEL} ({e}); returning a ZERO vector — "
            "semantic ranking is degraded until this is resolved."
        )
        return [0.0] * EMBED_DIM

class EpisodicMemory:
    def __init__(self, persist_dir: str):
        # ITEM: Cross-session persistence confirmed
        # Real ChromaDB persistent client — data survives app restarts
        if chromadb is None:
            logging.error("chromadb is not installed. Episodic memory will not work.")
            self.client = None
            return

        logging.getLogger("chromadb").setLevel(logging.ERROR)
        self.client = chromadb.PersistentClient(
            path=persist_dir,
            settings=Settings(anonymized_telemetry=False)
        )

    def _get_collection(self, persona: str):
        if not self.client:
            return None
        return self.client.get_or_create_collection(f"persona_{persona.lower()}")

    async def add(self, text: str, metadata: dict, persona: str = "jarvis"):
        # ITEM: Memory write — EXE (exact add() call with collection=f"persona_{active_persona}")
        collection = self._get_collection(persona)
        if not collection:
            return
            
        embedding = await embed(text)
        metadata["persona"] = persona.lower()
        metadata["pinned"] = metadata.get("pinned", False)
        
        def _add():
            collection.add(
                documents=[text],
                embeddings=[embedding],
                metadatas=[metadata],
                ids=[str(uuid4())]
            )
        await asyncio.to_thread(_add)

    async def query(self, text: str, top_k=5, persona: str = "jarvis") -> list[dict]:
        # ITEM: Memory read/query — EXE (exact query() call filtered to active persona collection)
        collection = self._get_collection(persona)
        if not collection:
            return []
            
        embedding = await embed(text)
        
        def _query():
            return collection.query(
                query_embeddings=[embedding], 
                n_results=top_k
            )
            
        results = await asyncio.to_thread(_query)
        
        out = []
        if results and results.get("documents") and results["documents"][0]:
            docs = results["documents"][0]
            metas = results["metadatas"][0] if results.get("metadatas") and results["metadatas"][0] else [{}] * len(docs)
            ids = results["ids"][0]
            for d, m, i in zip(docs, metas, ids):
                out.append({"id": i, "text": d, "metadata": m})
        return out

    async def pin(self, memory_id: str, persona: str = "jarvis"):
        # ITEM: Memory pin — EXE (exact metadata update to mark pinned=True)
        collection = self._get_collection(persona)
        if not collection:
            return
            
        def _pin():
            result = collection.get(ids=[memory_id])
            if result and result['metadatas'] and len(result['metadatas']) > 0:
                meta = result['metadatas'][0]
                meta['pinned'] = True
                collection.update(
                    ids=[memory_id],
                    metadatas=[meta]
                )
        await asyncio.to_thread(_pin)

    async def delete(self, memory_id: str, persona: str = "jarvis"):
        # ITEM: Memory delete — EXE (exact delete() call)
        collection = self._get_collection(persona)
        if not collection:
            return
            
        def _delete():
            collection.delete(ids=[memory_id])
        await asyncio.to_thread(_delete)

    async def embed_model_description(self, description: str, model_path: str, persona: str):
        collection = self._get_collection(f"{persona}_models")
        if not collection:
            return
        def _add():
            collection.add(
                documents=[description],
                metadatas=[{"model_path": model_path, "engine": "stable_fast_3d_local"}],
                ids=[str(uuid4())]
            )
        try:
            await asyncio.to_thread(_add)
        except Exception as e:
            logging.error(f"ChromaDB embed error: {e}")

    async def query_similar_model(self, description: str, threshold: float = 0.85, persona: str = "JARVIS") -> dict | None:
        collection = self._get_collection(f"{persona}_models")
        if not collection:
            return None
        def _query():
            return collection.query(
                query_texts=[description],
                n_results=1
            )
        try:
            results = await asyncio.to_thread(_query)
            if results and results.get('distances') and results['distances'][0]:
                dist = results['distances'][0][0]
                if dist <= (1.0 - threshold):
                    meta = results['metadatas'][0][0]
                    return {"model_path": meta.get("model_path")}
        except Exception as e:
            logging.error(f"ChromaDB query error: {e}")
        return None