[hongchuwudi]

Closes #1194

Summary

teamai import collected at most 200 source files per repository from a hardcoded cap, and truncated silently — a repository with more files produced a partial knowledge graph with no signal. This adds --max-files <n> to set the cap per run (the default stays 200) and warns with both counts when a source is truncated.

Change

• CLI: teamai import --max-files <n>, validated as a positive integer (--max-files must be a positive integer otherwise, exit 1).

• Plumbing: import.ts → --dir / --from-repo / --from-repo-list → importLocalCheckout / importFromRepo / importFromRepoList → importCodebase → extractCodebase. The extractor already accepted maxFiles; only the CLI entry was missing.

• Warning: collectCode now also returns the candidate count, and extractCodebase warns when files were dropped instead of truncating silently:

  [extract] demo-repo: 6 source files matched, but only the first 3 were collected. Raise the limit with --max-files <n>.
  

  Collection order is unchanged (key files first).

• Docs: docs/guide/knowledge.md and its zh-CN page; skill-data/core/references/commands.md regenerated with npx vitest run commands-reference -u.

Type of Change

• New feature (non-breaking change that adds functionality)

Test Plan

• npm run build ✓, npx tsc --noEmit ✓, npm run lint ✓ (0 errors).
• npx vitest run codebase-extract-fallback → 18 passed, including two new tests: --max-files caps collection and warns with both counts; no warning when every candidate fits.
• Affected suites (codebase-*, import-*, ci-extract-mr, anchors, dry-run-load-path, scala/swift-extractor): 244 passed, 5 failed — all five are pre-existing Windows-only failures. A run on pristine origin/main (fe3b906) reproduces the identical failures (C:/ vs C:\ path forms in anchors, a | in a temp path in codebase-reconcile, EPERM/timeout in dry-run-load-path).
• Real CLI run from dist/index.js against a temp repo with 6 source files:

$ teamai import --help | grep max-files
  --max-files <n>  Max source files collected per repository (default 200)

$ teamai import --dir <repo> --output <out> --skip-enrich --max-files 3
[extract] demo-repo: 6 source files matched, but only the first 3 were collected. Raise the limit with --max-files <n>.
[extract] demo-repo complete
  Files: 3
  Facts: 3 (component:3)
  ...
✔ Local directory demo-repo import complete

$ teamai import --dir <repo> --output <out2> --skip-enrich      # default 200
[extract] demo-repo complete
  Files: 6
  ...

$ teamai import --dir <repo> --output <out3> --skip-enrich --max-files 0
✖ --max-files must be a positive integer            (exit 1)

$ teamai import --dir <repo> --output <out4> --skip-enrich --max-files abc
✖ --max-files must be a positive integer            (exit 1)

Notes for Reviewers

• The default intentionally stays at 200. The issue also floated raising it, but that changes collection cost for every existing import, so I left that as a separate decision.
• --from-repo-list forwards the same cap to each repository it imports. --from-mr and teamai ci extract-mr keep their own internal limits and are untouched.

[github-actions]

No findings in the scoped PR diff.

The PR description documents sufficient testing, including a representative real-CLI run covering truncation, default collection, and invalid --max-files values.

Review was read-only; I did not run, build, or install anything from the PR.