{
  "markdown": "# vaas-x\n\n[![PyPI](https://img.shields.io/pypi/v/vaas-x.svg)](https://pypi.org/project/vaas-x/)\n[![Python](https://img.shields.io/pypi/pyversions/vaas-x.svg)](https://pypi.org/project/vaas-x/)\n[![License](https://img.shields.io/badge/license-proprietary-lightgrey.svg)](LICENSE)\n\nPersistent, outcome-grounded memory for AI agents and connected devices.\n\n`pip install vaas-x` gives any device or agent a persistent, queryable\nrecord of what it's seen — what happened, what it did, what the outcome\nwas — and retrieves the closest matching past episode in milliseconds, on\nCPU, with no GPU and no vector database to stand up yourself.\n\n## Install\n\n```bash\npip install vaas-x\n```\n\nGet a free-tier API key at [vaasx.com](https://vaasx.com) — no card\nrequired.\n\n## Quickstart\n\n```python\nfrom vaasx import Bootstrap\n\nbrain = Bootstrap(api_key=\"...\", device_id=\"my_device\")\n\n# Point it at a structured data stream (sensor readings, agent actions,\n# API events -- anything JSON-shaped) and it profiles the stream\n# automatically, no schema design required.\nbrain.connect(\"https://your-data-stream.com/events\")\n\n# Query for the most similar past episodes, ranked by what actually\n# worked, not just what's nearest in vector space.\nhits = brain.query(\"engine running hot\", k=5, prefer_success=True)\n```\n\nAn agent that already knows its own outcome the moment it acts can log a\nfull state/action/outcome episode in one call instead of a separate\ningest-then-tag-later round trip:\n\n```python\nresp = brain.ingest([{\n    \"state\": {\"observation\": observation},\n        \"action\": {\"taken\": action},\n            \"outcome\": {\"success\": result.success},\n            }])\n            episode_id = resp[\"episode_ids\"][0]\n\n            # Report back on a retrieved episode once you know how it went\n            brain.outcome(hits[0][\"id\"], success=True, delta=0.12)\n            ```\n\n            Three calls — `ingest`, `query`, `outcome` — cover the whole loop. No\n            schema to design, no vector DB to host, no embedding model to serve\n            yourself.\n\n            ## What it does\n\n            - Profiles a raw data stream automatically — statistical profiling,\n              schema classification, and anomaly detection all run locally before\n                anything touches the network.\n                - Stores episodes as (state, action, outcome) triples, so retrieval can be\n                  weighted toward what worked last time in a similar situation, not just\n                    nearest by embedding distance.\n                    - Runs on both a small edge device and a standard server — same codebase,\n                      no domain-specific build. Verified live on a £30 ESP32 microcontroller\n                        and on a standard EC2 instance.\n\n                        ## Numbers, not adjectives\n\n                        - 14ms mean retrieval across 1.18M stored episodes on a standard AWS\n                          m7i-flex.large CPU instance — no GPU, no AVX-512.\n                          - Blind-tested on NASA's CMAPSS FD001 predictive-maintenance benchmark:\n                            given 24 sensor channels with every identifier stripped, correctly\n                              classified all 24 with zero prior domain knowledge.\n                              - Consistent zero-config classification result across 8 independent\n                                subjects on a real wearable IMU dataset (UCI Daily and Sports\n                                  Activities).\n\n                                  Every number above is independently reproducible — see below.\n\n                                  ## Reproduce the benchmarks yourself\n\n                                  Every result this SDK's marketing points to has a step-by-step guide\n                                  against a real public dataset or environment, with your own API key, plus\n                                  an independent local cross-check that doesn't depend on the SDK at all:\n\n                                  - [CMAPSS predictive maintenance](https://vaasx.com/guides/cmapss.html)\n                                  - [Room-occupancy detection](https://vaasx.com/guides/occupancy.html)\n                                  - [Memory-augmented CartPole](https://vaasx.com/guides/cartpole.html) (zero download)\n                                  - [Wearable activity classification](https://vaasx.com/guides/sports_wearable.html)\n                                  - [Physics simulations](https://vaasx.com/guides/scientific_computing.html) (QEC, lattice field theory, Ising model)\n                                  - [Live drone control loop](https://vaasx.com/guides/drone_control.html)\n\n                                  Full guide index: [vaasx.com/guides](https://vaasx.com/guides)\n\n                                  ## Tiers\n\n                                  Free gets you an API key instantly. Developer (£19/mo) and Professional\n                                  (£99/mo) are self-serve paid tiers with higher episode limits and\n                                  additional capabilities. Enterprise is contract-based, for teams wanting\n                                  on-prem or air-gapped deployment. See\n                                  [vaasx.com/pricing](https://vaasx.com/pricing) for current details.\n\n                                  ## The honest caveat\n\n                                  This package ships as a compiled wheel, not source-available — the\n                                  retrieval and indexing internals are proprietary and protected as a trade\n                                  secret. What *is* open: full API documentation, working examples, and the\n                                  wire format, so you can see exactly what goes in and comes back even\n                                  without reading the internals.\n\n                                  Multimodal support (vision/audio/depth/thermal, via the optional `ac`\n                                  extra) is real but still early — encoder models are in active training\n                                  and not yet production-tuned; treat it as a preview, not a finished\n                                  feature.\n\n                                  ## Issues & questions\n\n                                  This repo doesn't carry the source (see above), but it's the right place\n                                  to file bugs against the published package, ask integration questions, or\n                                  request a reproduction guide for a dataset/environment not covered above.\n\n                                  ## Links\n\n                                  - [Documentation](https://vaasx.com/docs)\n                                  - [Whitepapers](https://vaasx.com/whitepapers) — technical deep dives, one per use case\n                                  - [Homepage](https://vaasx.com)\n\n                                  ## License\n\n                                  Proprietary — see [LICENSE](LICENSE).\n",
  "bytes": 7021,
  "sha": "a3b4db33c44a284692596e81514a82a1c29dca333d0109886a9ac27cb1db7a98",
  "repo_slug": "vaas-x/vaas-x",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_io_github_vaas_x_vaasxmcp_e443e752/readme"
}