Five improvements that eliminate most of the agent-in-the-loop friction
observed while closing out the 988-entity WoN evaluation (C.1):
1. Gemini adapter now retries on 429 + 5xx with exponential backoff
(same pattern already used by OpenRouter/OpenAI). Removes the need
for shell-level retry wrappers when hitting free-tier rate limits.
2. evaluate CLI prints the underlying error ("ERROR — HTTP 503 …")
instead of a bare "ERROR", so agents don't have to drop into Python
to diagnose transient failures.
3. --entity/--chapter now respect existing evaluation files by default
(previously only the full-collection pass did). New --force flag
opts into re-evaluation. Stops silently burning free-tier quota on
re-runs of the same slug.
4. --entity accepts hyphenated slugs (matching entity filenames) and
normalizes them to the underscore form used on disk. On a miss the
CLI suggests near matches instead of a bare "not found".
5. eval-summary --update-metrics is no longer destructive:
read_metrics_file/write_metrics_file preserve structured values
(type_distribution) and don't flatten ints to floats. Fixes a
silent data loss observed on every run.
Bonus: the evaluator field in written evaluation frontmatter now
falls back from run_config.model_name to the adapter's resolved model
(or the model echoed back in the API response), so rows no longer
show `evaluator: null` when --model is omitted.
Tests: new tests/unit/llm/test_gemini.py covers retry behavior;
tests/unit/infospace/test_history.py gains a round-trip test that
pins the type_distribution / int-preservation invariants.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Stage 1 — Decouple:
- Move RunConfig + LLMResponse to markitect/llm/models.py (canonical)
- Move LLMAdapter + Mock/ErrorLLMAdapter to markitect/llm/adapter.py
- markitect/prompts/execution/models.py and llm_adapter.py become re-export shims
- All 4 adapters + factory.py updated to import from markitect.llm.*
- Parameterize app_name in toml_config.py (resolve_llm, get_default_layers,
get_preference_layers): paths and env var now derived from app_name arg
- Add tests/test_llm_isolation.py: 7 isolation + backward-compat tests
Stage 2 — Extract:
- Standalone llm-connect package created at ~/llm-connect/
- All 18 llm files copied; markitect.* imports replaced with llm_connect.*
- LLMError base inlined in llm_connect/exceptions.py (no markitect dep)
- llm-connect installed into markitect-venv; declared in pyproject.toml
Smoke test: markitect llm-check succeeds (live Gemini API call).
Backward compat: markitect.prompts.execution.{models,llm_adapter} still work.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add GeminiAdapter calling Google's Generative Language REST API
(default model: gemini-2.5-flash). Register "gemini" as third
provider in the factory and CLI. Add rate-limit retry with
exponential backoff to the pipeline's _call_llm helper. Increase
default max_tokens from 2000 to 4096.
Process book-1-chapter-05 via Gemini free tier — 1 new entity
extracted (necessaries-conveniencies-and-amusements-of-life),
41 existing entities correctly skipped by dedup. Canonical set
now at 42 unique entities.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
