Quickstart¶
Get Argos running in five minutes. This assumes you already have Hermes installed. If not, see Installation for the full path.
1. Install the plugin¶
Copy argos_plugin/ into your Hermes plugins directory:
%LOCALAPPDATA%\hermes\plugins\hybrid_memory\ (Windows)
~/.hermes/plugins/hybrid_memory/ (Linux/macOS)
Restart Hermes. Pip dependencies (declared in plugin.yaml) install on first load:
duckdb,kuzu,sentence-transformers,fastapi,uvicorn,pydantic,pandas- Optional:
PyPDF2,openpyxl,python-docx(for structured ingestion)
2. Verify the tools loaded¶
hermes tools
You should see the memory_* tools listed. If they don't appear, check the Hermes log for import errors (usually a missing dependency).
3. Tell Argos something¶
In a Hermes conversation:
I live in Cape Town and I prefer dark mode for coding.
Argos extracts facts and creates proposals — they enter the review queue, they don't become active memory until you approve them.
4. Search what Argos remembers¶
What do you know about where I live?
Argos searches the store (vector + keyword fusion) and injects the relevant memories as context for the answer.
5. Review the proposal queue¶
Use the memory_candidate_list tool (or the CLI review tool) to see pending proposals. Approve the ones you want to keep; reject the rest. Approved proposals become active memories with full provenance.
6. Optional: start the external API¶
If you want non-Hermes agents to read your memory store:
# REST server (loopback-only, token-authenticated)
ARGOS_REST_TOKEN=<your-token> python -m argos_plugin.rest_server --home <hermes-home> --port 8732
# MCP server (stdio)
python -m argos_plugin.mcp_server --home <hermes-home>
See API reference for the full operation set.
Next steps¶
- Configuration — tune retrieval, embeddings, graph, and chains.
- API reference — MCP and REST endpoints.
- Tuning — embedder and reranker guidance.