ChronoMind

ChronoMind is a temporal memory retrieval system for personal knowledge. It treats memories as an evolving event graph and answers questions by combining semantic retrieval, graph traversal, lexical matching, temporal ranking, and timeline reconstruction.

This is not a chatbot. It is a search-and-reasoning engine for personal history.

What it does

• stores atomic memory events
• indexes memories in Qdrant for semantic search
• links events in Neo4j for causal and temporal traversal
• supports BM25 lexical retrieval for exact matches
• understands query intent before retrieval
• ranks results with temporal, causal, graph, and importance signals
• reconstructs a chronological context timeline
• generates an answer grounded in retrieved memory events

Core capabilities

• decision tracing
• belief evolution
• causal inference
• temporal evolution search
• comparison queries
• retrieval debugging and evaluation

Architecture

Layer	Role
Frontend	Next.js UI for query, timeline, graph, and debug inspection
Backend	FastAPI API for ingestion, retrieval, ranking, and reasoning
Vector search	Qdrant semantic retrieval
Graph search	Neo4j multi-hop traversal
Lexical search	BM25 exact-term matching
Ranking	Hybrid fusion, reranking, diversity filtering, temporal weighting
Context	Chronological reconstruction for LLM input

Repository layout

backend/       FastAPI app, retrieval pipeline, graph, vector, ingestion
frontend/      Next.js application and API routes
scripts/       Demo seeding utilities
evaluation/    Retrieval evaluation harness
docker-compose.yml  Full local stack

Quick start

1. Configure environment

cp .env.example .env

Fill in API keys if you want hosted embeddings or LLM reasoning:

• OPENAI_API_KEY
• ANTHROPIC_API_KEY

2. Start the stack

docker compose up -d --build

3. Seed demo data

python scripts/seed_demo.py

4. Open the app

• Frontend: http://localhost:3000
• Backend health: http://localhost:8000/health
• API docs: http://localhost:8000/docs
• Neo4j Browser: http://localhost:7474

Example queries

• What led me to change my tech stack?
• How has my opinion on remote work changed?
• What led to my ML engineering career?
• When did I first learn about machine learning?

API endpoints

Endpoint	Method	Description
/api/ingest	POST	Ingest raw text and extract atomic memory events
/api/query	POST	Run hybrid retrieval and answer a memory query
/api/timeline/{concept}	GET	Fetch a chronological timeline for a concept
/api/graph/explore	GET	Inspect the memory graph
/api/stats	GET	View corpus and graph statistics
/health	GET	Health check

Evaluation

Run the evaluation harness against the live API:

python evaluation/run_eval.py

It reports:

• Recall@K
• MRR
• NDCG@10
• Temporal Ordering Accuracy
• Redundancy Score

Demo focus

The seeded dataset is designed to show:

• learning a topic over time
• changing an opinion
• making a decision
• causal chains across events
• knowledge evolution through beliefs and refinements

Development notes

• The system runs fully in Docker Compose.
• The backend uses real retrieval logic end to end.
• The frontend exposes the retrieval trace for debugging and interview demos.

License

No license has been specified yet.