Tags: iarjunganesh/drift

v1.0.0 — Devpost submission complete

Freezes the final judge path and closes the Devpost Developer Tools
submission: demo video published (https://youtu.be/6sbIz0ZR8Hw, 2:45),
frame-by-frame verified against the shooting script; README, submission
docs, and Devpost-formatted mirror synced with no stale placeholders;
hosted API verified live at v0.10.2 (/health: ok, live, 0.10.2).

fix: deterministic upstream version-bump classification

Adds backend/core/versioning.py (deterministic version-tag diff, never a
model guess) and scripts/backfill_upstream_release_type.py; applied to
the three reviewed Insights that had no self-declared source statement
(JAX -> minor, NCCL -> patch, TensorRT -> minor). See ADR-012. Gates:
ruff, mypy, 189 tests at 100% backend coverage, frontend production
build.

release: VS Code MCP client evidence, demo assets, and chat grounding…

… fix

MCP client evidence for the deployed v0.10.0 API, the production demo
shooting script and tts-1 narration assets, and a /chat grounding fix: an
empty grounded_insight_ids (a genuine decline) no longer falls back to
citing every retrieved insight. Requires a redeploy for the fix to take
effect on the hosted app. Gates: ruff, mypy, 160 tests at 100% backend
coverage, frontend production build.

v0.10.0 — thin-client MCP integration over the public API (ADR-011)

release: v0.9.1 — Terra grounded-chat evidence pass

Bump app version 0.9.0 -> 0.9.1 across config, packaging, and lockfiles
(backend/core/config.py, pyproject.toml, frontend/package.json + lockfiles).

Evidence:
- Record the bounded Tier.LIVE (gpt-5.6-terra) grounded-chat capture against
  the reviewed store, with its Codex Session ID and SHA-256 integrity manifest
  (assets/evidence/2026-07-18-all-sources-terra.json).
- Refresh the README screenshot gallery into a numbered Luna -> Sol -> Terra
  flow: landing, API docs, Luna/Sol briefings with claim evidence, Ask DRIFT
  box, and two Terra grounded answers (NCCL, TensorRT).

Docs:
- Update CHANGELOG, README, and supporting docs to describe v0.9.1 as a
  source, evidence, and documentation release. Hosted app verification stays
  at v0.8.0 pending redeploy; the live store still serves the same five
  reviewed Tier.FINAL Insights.

feat: prepare v0.9.0 reviewed Sol evidence

Add the bounded Tier.FINAL reviewed evidence archive, auditable Luna retraction path, and explicit Sol/Luna notebook records. Synchronize release, deployment, session, submission, and screenshot documentation, add the planned v0.9.1 Terra run, and verify the repository with Ruff, Mypy, 155 tests at 100% coverage, and the frontend production build.

fix(chat): restrict citations to grounded insights

Make live Ask DRIFT responses structured so the model reports the Insight IDs it used and the API cites only those sources. Add malformed-response and subset-citation coverage, bump source and frontend metadata to 0.8.1, and synchronize release records while keeping hosted v0.8.0 verification explicit.

v0.8.0: publish verifiable fixture evidence and grounded chat

DRIFT v0.7.0: publish bounded reviewed capture

v0.6.1: theme-aware briefing and API branding