Skip to content

v0.2.0 — model-agnostic for real

Latest
Latest

Choose a tag to compare

View all tags
@jeremylongshore jeremylongshore released this 28 Aug 00:00
v0.2.0
a9b48f8
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Claude is no longer the only gated model: OpenAI (gpt-4o) passed the provider eval gate (all 7 golden fixtures, real-key run 2026-08-20) and joins SUPPORTED_PROVIDERS. Bring an OPENAI_API_KEY and the full pipeline runs on ChatGPT — no override flags.

Also in this release:

  • Vertical pack seam + compliance module: fail-closed DNC / TCPA quiet-hours / service-area gate runs before drafting; blocked contacts are recorded, never drafted. b2b-sdr ships as the default pack.
  • Plugin re-architecture: the /intent-outreach orchestrator skill dispatches per-phase sub-agents (researcher fan-out, enricher, drafter) over the bundled MCP server, plus companion slash-command skills and a connector-readiness hook.
  • Strict-structured-output-compatible seam schemas (required / nullable, no optionals) — portable to every provider's strict json_schema mode; this is what the first real OpenAI eval run caught.
  • Dependency refresh: Vercel AI SDK v6, zod v4.

Grok and Gemini adapters ship ready and gate in the same way: npx tsx evals/run.ts --providers xai (or google) with a real key. Full detail in the CHANGELOG.

v0.1.0 (the June 16 rebuild) is also now tagged retroactively at its true release commit.

Assets 2
Loading