DemandScope v0.3.0

Full Changelog: v0.2.0...v0.3.0

DemandScope v0.2.0

DemandScope v0.2.0 — release notes (draft, ready for GitHub release)

Date: 2026-08-17
One-liner: Dependency-free MCP server that lets any AI agent validate
product ideas with live public-API demand signals — now with trend deltas,
caching, and rate-limit backoff.

Highlights

• New tool: hn_trend — Hacker News story count for the last N days vs the
  prior window, with growth_ratio (attention acceleration).
• New tool: github_trend — GitHub repos created in the last N days vs the
  prior window, with growth_ratio (supply/interest acceleration).
• File-based TTL cache (DEMANDSCOPE_CACHE_DIR): GitHub/HN 1h, npm/PyPI
  24h. Repeat calls are free — protects upstream APIs and your rate limits.
• Exponential backoff on HTTP 429/5xx honoring Retry-After (max 4 tries);
  pypistats.org remains best-effort.
• Packaging: pyproject.toml (setuptools, demandscope console script),
  MIT license.
• Tests: 10/10 end-to-end protocol tests pass, including a cache-hit
  assertion against an isolated cache dir.

Compatibility

• Python >= 3.9, zero dependencies (stdlib only).
• MCP protocol 2025-06-18; implements initialize, ping, tools/list,
  tools/call over stdio.

Known limitations

• Download absolutes include CI/mirror traffic — directional only.
• GitHub search unauthenticated limit is 10 req/min; github_trend self-throttles
  (7s between its two window queries).

Publication status

Everything needed for gh repo create + release is in-repo: LICENSE (MIT),
README with AI-transparency notice, pyproject.toml, passing tests, these notes.
Single remaining step is owner-side GitHub authentication
(see ../../NEEDS_OWNER_ACTION.md, item 1).