[nightshift] test-gap: 4,130 lines across 12 modules lack test coverage

[nightshift-micr]

Test Coverage Gap Analysis

Module: Microck/traccia
Task: test-gap
Severity: Medium — 4,130 lines of untested source code across 12 modules

---

Summary

Only 5 of 17 source modules (29%) have dedicated test files. The remaining 12 modules containing 4,130 lines of code have zero test coverage. The existing test suite (62 tests across 8 files) covers parsing, pipeline orchestration, models, source detection, and CLI init commands well, but leaves critical business logic untested.

Coverage Map

Source Module	Lines	Has Tests	Public Symbols	Risk
rendering.py	930	✗	render_project, export_obsidian, ascii_tree, mermaid_tree	HIGH — Largest untested module, output correctness is critical
storage.py	517	✗	Storage class with 20+ methods	HIGH — Data persistence layer, all CRUD untested
llm.py	506	✗	extract_evidence, canonicalize, score_skill, healthcheck	HIGH — Core AI pipeline, no backend contract tests
family_normalizer.py	449	✗	normalize_family_content, rendered_text	MEDIUM — HTML normalization logic
cli.py	432	✗	26 CLI commands (init, lint, doctor, add, ingest, etc.)	MEDIUM — Partially tested via test_init.py integration tests
bootstrap.py	390	✗	RepoInitializer, initialize	MEDIUM — Project setup logic
document_normalizer.py	287	✗	normalize_document	MEDIUM — Document processing
extraction.py	192	✗	extract_evidence	MEDIUM — Evidence extraction
pipeline_support.py	163	✗	support_score, should_create_node, build_skill_node	MEDIUM — Skill graph construction
taxonomy.py	123	✗	DomainDefinition, SkillDefinition, match_skill_names	LOW — Schema definitions
config.py	96	✗	TracciaPaths, ThresholdConfig, load_config	LOW — Config loading
utils.py	45	✗	slugify, short_hash, file_sha256, skill_id	LOW — Utility functions

Existing Test Coverage (62 tests, 8 files)

Test File	Tests	What It Covers
test_pipeline.py	20	End-to-end pipeline: ingest, reingest, rendering, graph, manifests
test_parsers.py	13	Document parsing: chat, Instagram, Twitter, Reddit, DOCX, PDF
test_init.py	9	CLI init/doctor commands, OpenAI backend contract
test_signal_handling.py	7	Export classification, AI trace detection
test_source_detection.py	6	Family detection from paths and archives
test_npm_wrapper.py	4	NPM wrapper subprocess handling
test_models.py	2	Config schema and domain model validation
test_fixtures.py	1	Golden fixture loading

Priority Recommendations

1. rendering.py (930 lines, HIGH risk) — Add tests for render_project, export_obsidian, ascii_tree, mermaid_tree. These produce the user-facing output and any bug here is immediately visible. Test with various graph structures (empty, single-node, deep tree, cycle).

2. storage.py (517 lines, HIGH risk) — Add integration tests for the Storage class. Test CRUD operations, query patterns, and error handling for missing/corrupt data. Use an in-memory SQLite fixture or mock the database layer.

3. llm.py (506 lines, HIGH risk) — Add contract tests for extract_evidence, canonicalize, score_skill. Mock the HTTP backend to test parsing of valid/invalid/malformed LLM responses. The existing test_init.py tests some OpenAI backend handling but doesn't cover the full extraction→canonicalization→scoring pipeline.

4. pipeline_support.py (163 lines, MEDIUM risk) — Unit test support_score, should_create_node, build_skill_node. These are pure functions with deterministic logic that are easy to test in isolation.

5. utils.py (45 lines, LOW risk but easy win) — Test slugify, short_hash, file_sha256, skill_id with edge cases (empty strings, unicode, special characters). Quick to implement, high confidence gain.

Estimated Effort

Priority	Module	Estimated Tests	Effort
P0	utils.py	8-10	1 hour
P0	pipeline_support.py	10-12	2 hours
P1	rendering.py	15-20	4 hours
P1	storage.py	12-15	4 hours
P2	llm.py	12-15	3 hours
P2	family_normalizer.py	8-10	2 hours
P3	config.py, taxonomy.py, extraction.py	10-12	2 hours

---

Generated by nightshift — autonomous code quality bot.

[Microck]

Status after maintenance pass: partially improved, intentionally left open.

Addressed now:

• test count increased to 147 passing tests.
• added targeted coverage for config, utilities, extraction, storage identifier safety, LLM transport/header behavior, and error messages.
• CI now validates the expanded suite on Python 3.12 and 3.13.

Still open:

• large modules like rendering.py, pipeline.py, family_normalizer.py, and broader CLI behavior still need more dedicated coverage.
• this issue should remain open until coverage is measured and the highest-risk modules have meaningful regression tests.