Skip to main content

everalgo-knowledge

Extracts a flattened topic tree (list[KnowledgeMemory]) from parsed content (ParsedContent) via LLM.

Use

from everalgo.knowledge import KnowledgeExtractor
from everalgo.types import ParsedContent

parsed = ParsedContent(text="...", mime="text/markdown")
memories = await KnowledgeExtractor(llm=client).aextract(parsed, doc_id="doc1", title="My doc")
# memories[0] = synthetic doc root; rest = DFS-ordered topic nodes.

llm: any everalgo.llm.LLMClient, passed at construction. Every call sends a system message of rules and a user message of data, requests a strict reply model as response_format, and raises everalgo.llm.LLMOutputError when the reply does not validate (no retry; the host decides). Sync bridge: extract(...).

custom_instructions= (on aextract and aclassify_category) adds the host's text to the bundled rules, never replacing them: it goes, verbatim and never rendered as a template, in a <custom_instructions> block closing the system message of every LLM call the extraction makes; None or a blank string adds nothing.

Pipeline at a glance

ParsedContent
  → preprocess + atomize           (_block_split)
  → topic-tree LLM (per batch)     (prompts.en.topic_tree_extract)
  → cross-batch merge (>1 batch)   (prompts.en.content_merge + topic_merge)
  → postprocess: split unsplit leaves + assign uncovered blocks
  → tree assembly + DFS flatten    → list[KnowledgeMemory]

Install

pip install everalgo-knowledge    # pulls everalgo-core + everalgo-parser

For development within the monorepo workspace:

uv sync --package everalgo-knowledge

Live-LLM scripts

Run from the repository root. Put LLM_API_KEY, LLM_BASE_URL, LLM_MODEL in a .env file there first; each command block below sources it inline and is paste-and-run.

Smoke test — 2 fixtures, structural assertions only, no artifacts written:

set -a; source .env; set +a
uv run pytest -m integration packages/everalgo-knowledge/tests/functional/test_extraction_smoke.py

Run one doc — dumps result (JSON + optional HTML viz) under tests/functional/outputs/:

set -a; source .env; set +a
uv run python packages/everalgo-knowledge/examples/run_one_doc.py \
  packages/everalgo-knowledge/tests/functional/fixtures/idx_multi_topic.json --html
  • everalgo-parser — produces ParsedContent from raw files (fully implemented except video, which is deferred pending ADR)
  • everalgo-core — KnowledgeMemory type is defined here

Metadata

Release files for everalgo-knowledge 0.3.0

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

Source distribution (sdist)

Source distribution for everalgo-knowledge 0.3.0
File Size Uploaded Provenance Actions
everalgo_knowledge-0.3.0.tar.gz 73.7 kB Without provenance Details for everalgo_knowledge-0.3.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for everalgo-knowledge 0.3.0
File Interpreter ABI Platform Actions
everalgo_knowledge-0.3.0-py3-none-any.whl Python 3 none any Details for everalgo_knowledge-0.3.0-py3-none-any.whl

Total release size: 115.5 kB

Release files / everalgo_knowledge-0.3.0.tar.gz

Download URL everalgo_knowledge-0.3.0.tar.gz
Size 73.7 kB
Tags Source
SHA-256 checksum
How to use checksums
6d2df8d6105af485b1deddafd19674267608748951a54f0eb16f1b6322706204
BLAKE2b-256 checksum
How to use checksums
1d1822ad3c415b40cee8b1f53f9d799fffc380b357729080c285a450eaa22470
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.12.12

Release files / everalgo_knowledge-0.3.0-py3-none-any.whl

Download URL everalgo_knowledge-0.3.0-py3-none-any.whl
Size 41.8 kB
Tags Python 3
SHA-256 checksum
How to use checksums
3519a2acbe63f21fd9e2b88baae15e6aa84e88864c6036d4c39bf6f51c65b992
BLAKE2b-256 checksum
How to use checksums
f43ade0d09c4ba8cb42c040d48cd9569130ca5c052398a5b7286db4a5ee403e0
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.12.12

Security for everalgo-knowledge 0.3.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

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