Memory infrastructure for the Hero stack — collections, conversion, Q&A, ontology extraction, embeddings, retrieval
- Rust 68.6%
- HTML 28.1%
- Shell 2.1%
- JavaScript 0.9%
- CSS 0.3%
|
|
||
|---|---|---|
| .forgejo/workflows | ||
| crates | ||
| docs | ||
| .gitignore | ||
| build.sh | ||
| Cargo.toml | ||
| Cargo.toml.hero_builder_backup | ||
| download_models.sh | ||
| favicon.svg | ||
| install.sh | ||
| MACOS_ONNX_FIX.md | ||
| MAKEFILE_ROBUSTNESS.md | ||
| OAUTH_DEBUG.md | ||
| openrpc.json | ||
| PURPOSE.md | ||
| README.md | ||
Hero Memory
Hero Memory is the memory infrastructure for the Hero stack. A single service that ingests documents and code, organises them into collections, extracts Q&A pairs and ontological graph structure, generates embeddings, and serves retrieval to agents.
It owns the pipeline from raw bytes to retrievable memory. It is not a
database (hero_db stores graph rows it produces) and it is not a model
server (hero_aibroker answers LLM calls; hero_memory_inference runs
the embedding models). It is the layer in between.
Documentation
- PRD — product requirements: what hero_memory does and the interfaces it exposes.
- ADRs — architecture decisions, with rationale.
Status
Specification phase. Implementation tracked via Forgejo issues against this repository.