updates — planner (JG2HQZWB)
0.1.6 — corpus-driven plans
Added data/corpus.json with exemplar projects (rubik solver, nd CLI). Planner now matches idea to nearest exemplar by keywords and uses its file tree, milestones, libraries, risks, and pitfalls. Context-driven non-LLM plans.
0.1.5 — expanded cli template
Expanded cli-tool template with modules (cli, config, error) and integration test. Template now compiles and tests pass.
0.1.4 — context-dependent base plan
Improved non-LLM plan: idea-aware features (keyword extraction), expanded file tree, context-dependent risks/pitfalls, cleaner feature normalization, better architecture.
0.1.3 — feature cleanup
Fixed feature parsing (JSON + strict validation), improved base plan with idea-aware libraries/milestones/risks/pitfalls, constraints support, file tree from template.
0.1.2 — fix feature leakage
Fixed refine_features to use JSON output with strict validation and fallback. Raised default max_tokens to 16384. Fixed risk-merge off-by-two. Improved feature suggestion parsing.
0.1.1 — fix feature handling
Fixed refine_features to validate output and fall back instead of storing chain-of-thought. Added --constraints flag, dedicated Constraints section, improved LLM output handling, raised default max_tokens to 8192. Added README.
v0.1.0
first release