Problem
Certain code.json fields, such as long description and tags, require non deterministic, generative content that the tool's current deterministic generation logic can't produce so they stay empty waiting for manual entry.
Goal
Add AI generation so the tool can fill in these fields, producing more filled out code.json metadata while holding to these design goals:
- AI-agnostic: not tied to any single AI provider
-Zero hosting cost: adding this can't introduce any cost/host burden
Constraints
- No self-hosted AI: the tool's own code cannot host or run an AI model.
- No hardcoded AI provider: the tool's code can make the API call itself, but the endpoint, credentials, and model are entirely supplied by the user somehow so that no specific AI vendor is built into the tool.
Context on inputs (not a proposed mechanism)
For fields like these, the AI would need context to work from. The repo's file tree and existing docs, think README and CONTRIBUTING, could be the kind of input that could supply context.
Problem
Certain code.json fields, such as long description and tags, require non deterministic, generative content that the tool's current deterministic generation logic can't produce so they stay empty waiting for manual entry.
Goal
Add AI generation so the tool can fill in these fields, producing more filled out code.json metadata while holding to these design goals:
-Zero hosting cost: adding this can't introduce any cost/host burden
Constraints
Context on inputs (not a proposed mechanism)
For fields like these, the AI would need context to work from. The repo's file tree and existing docs, think README and CONTRIBUTING, could be the kind of input that could supply context.