Agent task: bootstrap NLP lab maturity spine for Holmes

[mdheller]

Agent-directed work package.

Target agent: GitHub Copilot coding agent; Codex may also take this if enabled via GitHub Agent HQ.

Context:
nlplab is the Linux-native lab workspace for classical NLP, linguistic primitives, entity/relation extraction, topic modeling, semantic parsing, and language pipeline evaluation for Holmes. It is a lab repo: it may experiment, evaluate, package evidence, and emit functional service manifests. It must not self-promote models into production.

Scope:

1. Inspect the live repository before editing.
2. Add a real README explaining the lab purpose and relationship to SocioProphet/holmes, SocioProphet/functional-model-surfaces, SociOS-Linux/embeddinglab, SociOS-Linux/translationlab, and SocioProphet/model-governance-ledger.
3. Add repo.maturity.yaml conforming to repo-maturity.v1 semantics.
4. Add lab.manifest.json describing supported NLP surfaces: tokenization, sentence splitting, POS/morphology, NER, relation extraction, classification, topic modeling, semantic parsing, and graph conversion.
5. Add service-manifest/functional-service.v1.json conforming to the functional service manifest schema.
6. Add examples/, evals/, datasets/README.md, adapters/README.md, training-runs/README.md.
7. Add Makefile validation target.

Acceptance criteria:

• make validate passes.
• Repo clearly states lab-only boundary: no production promotion, no secrets, no model blobs unless explicitly release-managed later.
• The functional service manifest can be consumed by functional-model-surfaces and model-governance tooling.
• Holmes can use this repo as its classical/neural NLP lab substrate.

Boundary:
One PR only. Do not touch unrelated repos. Do not add large model weights or datasets.

[mdheller]

@codex Please take this issue in the nlplab Codex environment. Inspect the repo, implement the bounded lab maturity spine described in the issue, keep it lab-only, open one PR, and include make validate evidence.

[mdheller]

@copilot Please take this issue if Copilot coding agent is available. Implement the bounded lab maturity spine, keep it lab-only, open one PR, and include make validate evidence.

[chatgpt-codex-connector]

Summary

• Replaced the placeholder README with a full lab charter that defines nlplab purpose, explicitly codifies the lab-only boundary (no production promotion, no in-repo secrets, no unmanaged model blobs), and documents relationships to Holmes ecosystem repos. README.mdL1-L59
• Added repo.maturity.yaml with repo-maturity.v1 semantics, bounded maturity metadata, integration links, and explicit non-production guardrails. repo.maturity.yamlL1-L25
• Added lab.manifest.json declaring all requested NLP capability surfaces (tokenization through graph conversion) and lab-only constraints. lab.manifest.jsonL1-L22
• Added service-manifest/functional-service.v1.json defining a functional service manifest with interfaces for each requested surface plus governance/compatibility metadata for downstream consumption. service-manifest/functional-service.v1.jsonL1-L59
• Added scaffold docs for datasets/, adapters/, training-runs/, examples/, and evals/ to establish expected lab artifacts and policies. datasets/README.mdL1-L9 adapters/README.mdL1-L9 training-runs/README.mdL1-L12 examples/README.mdL1-L3 evals/README.mdL1-L3
• Added a Makefile validate target that verifies required files, parses JSON manifests, enforces lab-boundary keys, and checks required NLP surface declarations. MakefileL1-L17

Testing

• ❌ make validate (initial run failed due to a Python one-liner syntax issue; fixed immediately in Makefile).
• ✅ make validate

Also completed follow-up workflow requirements:

• Committed changes on the current branch (3ca9c75).
• Opened a PR via make_pr with summary + validation evidence.

View task →