Skip to main content

awbrain

Your history as a wiki of linked markdown — every claim pinned to the file and line that produced it.

Every AI memory system sells you a black box: sessions vanish into a vendor cloud, and "it remembers" means "the vendor's model was prompted with a blob". Memory should be files you can read, claims you can check against the session or file that produced them, and retrieval that never dumps your whole history into a prompt.

awbrain is a folder of linked markdown you can open in any editor: no proprietary blob, no vendor cloud. Harvest your notes and sessions into a wiki, ask questions against only the top-k pages (never everything), and walk any claim back to the source line that supports it.

Install

pip install awbrain

Use

# build (or refresh) the wiki from a folder of notes/sessions
awbrain harvest ~/notes

# answer from ONLY the top-k pages — never your whole history
awbrain ask "what did we decide about the pool?"

# walk a claim back to the file and line that supports it
awbrain verify "a self-owned platform beats a subscription computer"

# re-harvest changed sources so the brain stays current
awbrain watch ~/notes

# receipts: sources, pages, claims, embedder mode
awbrain status

How it works

  • harvest reads a folder of markdown/notes and writes a linked wiki: pages with [[wikilinks]], each claim pinned to its source file and line span.
  • ask retrieves top-k pages plus a two-hop wikilink expansion — never the whole history — and answers with [[citation]] markers.
  • verify walks a claim back to the source file and line span that supports it, or says it cannot find the evidence.
  • watch re-harvests changed sources so the brain stays current as work changes.

Embeddings come from a router (nomic-embed-text) when reachable, else an honest lexical fallback — the manifest records which mode was used, so the retrieval story is never overstated.

Compose with

  • awgraph — the wikilinks ARE a graph; ask already walks two hops.
  • awm — claims as scoped memory.
  • awseal — failures as "what not to do" claims.

MIT licensed.

Metadata

Release files for awbrain 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for awbrain 0.1.0
File Size Uploaded
awbrain-0.1.0.tar.gz 25.0 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for awbrain 0.1.0
File Interpreter ABI Platform
awbrain-0.1.0-py3-none-any.whl Python 3 none any Details

Total release size: 50.0 kB

Release files / awbrain-0.1.0.tar.gz

Download URL awbrain-0.1.0.tar.gz
Size 25.0 kB
Tags Source
SHA-256 checksum
How to use checksums
9e718aa93b7f50d2440219bdbbc91100bf437c666f36863f7a0aef418c948c37
BLAKE2b-256 checksum
How to use checksums
04cf69643743d2063ef8c46a6278149e714b506e088590f3133fa41a785a68df
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.12.3

Release files / awbrain-0.1.0-py3-none-any.whl

Download URL awbrain-0.1.0-py3-none-any.whl
Size 25.0 kB
Tags Python 3
SHA-256 checksum
How to use checksums
16c36e613c9a68aced4bf64d24afbaaf246973910a0711c672e0bdab4a700b81
BLAKE2b-256 checksum
How to use checksums
d2deead098d321eedb206aec1937015c765783ea5ffc2035df47eb2f9ed2065f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.12.3

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page