1caf5ec9ee
* feat: add JinaMLXCrossEncoder for native Apple Silicon reranking Adds a new `jina-mlx` reranker provider backed by jinaai/jina-reranker-v3-mlx, a 0.6B multilingual listwise reranker running via the MLX framework on Apple Silicon. The model is downloaded automatically from HuggingFace Hub on first use. Benchmarked latencies (Apple Silicon): 1 doc→32ms, 5→45ms, 10→60ms, 20→94ms. Sub-linear scaling because all docs are ranked in a single forward pass. - Embeds the MLX reranker implementation (_MLXReranker / _MLPProjector) directly in cross_encoder.py with no transformers/PyTorch dependency - Adds `mlx`, `mlx-lm`, `safetensors` to pyproject.toml optional deps (uv add) - Updates configuration.md with provider docs and benchmark table * refactor: import MLXReranker from repo rerank.py instead of duplicating code Use importlib to load MLXReranker directly from the model repo's own rerank.py (downloaded via snapshot_download). Also pin exact minimum versions for mlx>=0.31.0, mlx-lm>=0.31.1, safetensors>=0.6.2 (verified against installed versions). * refactor: move MLX reranker impl to dedicated jina_mlx_reranker.py Replaces the importlib hack with a proper module. jina_mlx_reranker.py is adapted from jinaai/jina-reranker-v3-mlx/rerank.py (CC BY-NC 4.0) with the source clearly documented at the top of the file. * docs: simplify jina-mlx reranker docs * fix: disable GIN fastupdate on source_memory_ids index to prevent deadlocks GIN fastupdate buffers inserts in a pending list and flushes it with AccessExclusiveLock when full. Under concurrent test load (8 xdist workers all running retain_async), two workers can trigger a flush simultaneously and deadlock. Recreating the index with fastupdate=off eliminates the flush/lock cycle at the cost of slightly slower individual inserts. * fix: drop per-bank HNSW indexes after transaction to avoid AccessExclusiveLock deadlock When deleting a bank, the previous code dropped HNSW indexes inside the same transaction as the DELETE FROM memory_units. Since DROP INDEX needs AccessExclusiveLock on the parent table and DELETE holds RowExclusiveLock, two concurrent bank deletions deadlocked on the same table lock. Fix: capture internal_id inside the transaction, commit, then drop the indexes outside the transaction so no row-level locks are held. |
||
|---|---|---|
| .. | ||
| hindsight_api | ||
| tests | ||
| pyproject.toml | ||
| README.md | ||
Hindsight API
Memory System for AI Agents — Temporal + Semantic + Entity Memory Architecture using PostgreSQL with pgvector.
Hindsight gives AI agents persistent memory that works like human memory: it stores facts, tracks entities and relationships, handles temporal reasoning ("what happened last spring?"), and forms opinions based on configurable disposition traits.
Installation
pip install hindsight-api
Quick Start
Run the Server
# Set your LLM provider
export HINDSIGHT_API_LLM_PROVIDER=openai
export HINDSIGHT_API_LLM_API_KEY=sk-xxxxxxxxxxxx
# Start the server (uses embedded PostgreSQL by default)
hindsight-api
The server starts at http://localhost:8888 with:
- REST API for memory operations
- MCP server at
/mcpfor tool-use integration
Use the Python API
from hindsight_api import MemoryEngine
# Create and initialize the memory engine
memory = MemoryEngine()
await memory.initialize()
# Create a memory bank for your agent
bank = await memory.create_memory_bank(
name="my-assistant",
background="A helpful coding assistant"
)
# Store a memory
await memory.retain(
memory_bank_id=bank.id,
content="The user prefers Python for data science projects"
)
# Recall memories
results = await memory.recall(
memory_bank_id=bank.id,
query="What programming language does the user prefer?"
)
# Reflect with reasoning
response = await memory.reflect(
memory_bank_id=bank.id,
query="Should I recommend Python or R for this ML project?"
)
CLI Options
hindsight-api --help
# Common options
hindsight-api --port 9000 # Custom port (default: 8888)
hindsight-api --host 127.0.0.1 # Bind to localhost only
hindsight-api --workers 4 # Multiple worker processes
hindsight-api --log-level debug # Verbose logging
Configuration
Configure via environment variables:
| Variable | Description | Default |
|---|---|---|
HINDSIGHT_API_DATABASE_URL |
PostgreSQL connection string | pg0 (embedded) |
HINDSIGHT_API_LLM_PROVIDER |
openai, anthropic, gemini, groq, ollama, lmstudio |
openai |
HINDSIGHT_API_LLM_API_KEY |
API key for LLM provider | - |
HINDSIGHT_API_LLM_MODEL |
Model name | gpt-4o-mini |
HINDSIGHT_API_HOST |
Server bind address | 0.0.0.0 |
HINDSIGHT_API_PORT |
Server port | 8888 |
Example with External PostgreSQL
export HINDSIGHT_API_DATABASE_URL=postgresql://user:pass@localhost:5432/hindsight
export HINDSIGHT_API_LLM_PROVIDER=groq
export HINDSIGHT_API_LLM_API_KEY=gsk_xxxxxxxxxxxx
hindsight-api
Docker
docker run --rm -it -p 8888:8888 \
-e HINDSIGHT_API_LLM_API_KEY=$OPENAI_API_KEY \
-v $HOME/.hindsight-docker:/home/hindsight/.pg0 \
ghcr.io/vectorize-io/hindsight:latest
MCP Server
For local MCP integration without running the full API server:
hindsight-local-mcp
This runs a stdio-based MCP server that can be used directly with MCP-compatible clients.
Key Features
- Multi-Strategy Retrieval (TEMPR) — Semantic, keyword, graph, and temporal search combined with RRF fusion
- Entity Graph — Automatic entity extraction and relationship tracking
- Temporal Reasoning — Native support for time-based queries
- Disposition Traits — Configurable skepticism, literalism, and empathy influence opinion formation
- Three Memory Types — World facts, bank actions, and formed opinions with confidence scores
Documentation
Full documentation: https://hindsight.vectorize.io
License
Apache 2.0