[satyan2142]

Adds KL4A to the catalog.

What it is: KL4A ingests SOP, policy and regulation documents - PDF, DOCX or plain text, prose written for a human reader rather than a structured corpus - and emits OKF v0.2 bundles. Every extracted claim carries the exact byte range of the source sentence it came from, and nothing is marked verified until a person approves it. Approvals, rejections and edits are retained as events in the bundle, each with a reviewer and a rationale.

Why it belongs here: the existing conversion entries convert structured or already-Markdown sources (binder takes a plain-markdown corpus, convert-to-okf takes awesome lists, JSON and URL lists). KL4A covers the unstructured-document end of the same problem. I filed it under conversion rather than opening a new document section for a single entry, since from the catalog's point of view it is a source-to-OKF converter and belongs next to its peers.

Apache-2.0. Ships a desktop app (Tauri, Windows/macOS/Linux) plus sopkb-cli and sopkb-mcp, a read-only stdio MCP server, as single binaries with every release. A worked example bundle lives at examples/glp1-healthcare if you want to read real output before installing anything.

Checklist

• Link resolves to the canonical repo
• Description is accurate, no marketing language
• Actively maintained, explicit OKF v0.2 support
• No duplicate: searched the catalog for kl4a, knowledge layer, cogniswitch
• Not a paid or spam listing

Validation

awesome-okf validate   -> OK - 33 entries validated
awesome-okf readme     -> Wrote README.md
pytest                 -> 9 passed

---

One note in case it is useful rather than as part of this PR: pip install -e ".[dev]" currently fails because pyproject.toml requires okf>=1.6.0,<2, and PyPI only has okf 0.1.0. Nothing under src/ imports okf, so I installed pyyaml, typer and rich directly and the CLI and tests ran clean. Happy to open a separate issue if you would like it tracked.

[Albertchamberlain]

The upstream KL4A repository, Apache-2.0 license, releases, and read-only MCP implementation are real, and the submitted English catalog row is structurally valid. Changes are required before merge:

1. Only README.md was regenerated. Awesome-OKF requires all localized READMEs to stay in sync; regenerate and commit EN/ZH/JA/KO (and RU if #18 is now in the base branch).
2. KL4A's mine flow defaults to an Azure LLM provider. The current catalog text reads like a local document converter but does not disclose that SOP/policy/regulation document content may be sent to an external model unless an offline fixture/provider is explicitly selected. State the default cloud processing, credential requirement, data egress, and how to select offline mode.
3. “Every claim carries the exact source span” and “nothing is marked verified until a human approves it” are strong guarantees; link the implementation/tests that enforce them or soften the wording.
4. For downloadable binaries, publish and recommend verification of checksums/signatures rather than relying on an unverified latest artifact.

Run the full catalog, README-set, and pytest checks after rebasing on current main.