Skip to content

feat(ai): switch chat model to deepseek-v4-flash - #322

Merged
fworks-tech merged 2 commits into
mainfrom
feat/319-deepseek-v4-flash
Sep 27, 2026
Merged

fworks-tech merged 2 commits into
mainfrom
feat/319-deepseek-v4-flash

Conversation

@fworks-tech

Copy link
Copy Markdown
Owner

What

Switches the chat assistant MODEL_ID from gpt-6-luna to deepseek-v4-flash in src/app/api/chat/route.ts.

Why

deepseek-v4-flash is cheaper than gpt-6-luna ($0.14/$0.28 vs $0.10/$0.50 per 1M tokens) and uses the standard chat/completions endpoint, avoiding the Responses API compatibility issues that caused 400 errors in production.

How to test

  1. npm run lint
  2. npm run typecheck
  3. npm test -- src/app/api/chat/tests/route.test.ts src/lib/abuse/tests/cost.test.ts

Fixes #319

Switch chat MODEL_ID from glm-5.3-flash to jev-1.13.

Closes #319
- Cheaper than gpt-6-luna: $0.14/$0.28 vs $0.10/$0.50 per 1M tokens
- Uses standard chat/completions endpoint (no Responses API needed)
- Update cost constant to match new rate
@fworks-tech
fworks-tech force-pushed the feat/319-deepseek-v4-flash branch from 930878a to 0886f57 Compare September 27, 2026 19:18
@fworks-tech
fworks-tech merged commit 326f10c into main Sep 27, 2026
2 of 3 checks passed
@fworks-tech
fworks-tech deleted the feat/319-deepseek-v4-flash branch September 27, 2026 19:18
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

feat(ai): use jev-1.13 model for AI assistant

1 participant