Open Model Review

Add Cost Review
to every pull request.

Route pull requests to open source models for a multi-dimensional review.

Try for free

Pull request review preview

IonWarpbotcommented 15 min ago

IonWarp Review — PR #412

3 issues| 2 P1 1 P2

#SevIssueFile

1Streamed turns record zero tokenslib/ai/chat.ts

2Ledger prices the requested model, not the served onelib/ai/ledger.ts

3Cache-read tokens not carried into usage rowlib/ai/usage.ts

lib/ai/chat.ts

575859

−++

const res = await client.chat.completions.create({ model, messages });const stream = await client.chat.completions.create({ model, messages, stream: true });for await (const chunk of stream) yield chunk.choices[0]?.delta?.content ?? "";

Streaming turned usage off · Cost Review

The switch to streaming never asks for usage (`stream_options: { include_usage: true }`), so every chat turn now writes 0 tokens and $0 to the ledger the budget reads. Request usage and record the final chunk's usage with the served model.

Suggested fixAI fix prompt

Pricing

Best pricing in the game.

One workspace price with seats included. Credits pool across your team. Unlimited repos on every plan.

Starter

$0 /mo

Try AI PR review.

Starter Swarm across 2 models
- Code Review
- Security Review
3 seats included
15,000 credits to start, then 5,000/mo
Unlimited repos

Get Started

Pro

Max

$149 /mo

Scale AI PR review.

Max Swarm across 5 models
- UX Review
- API Performance
- Analytics Review
- SEO Review
- Changelog
- CI/CD Performance
- Plan Review
- Repeat-Failure Learnings
- SOC 2 Review
10 seats included
+75,000 credits/mo
+$29/mo per added seat
Unlimited repos
Priority support

Choose Max

Enterprise

Custom swarms + credits · GitHub Enterprise · SSO/SAML + audit logs · Custom security review + SLA

IonWarp for Startups

Get the Max plan at half price. — Full Swarm. Proactive PRs.

Apply

How Cost Review works

Details

Runs on every pull request, on GitHub

Agent Prompt

Reviews a GitHub PR for AI usage and cost observability — a model call (direct, or through OpenRouter, an AI gateway or LiteLLM) whose record drops token usage, cost, the served model, the routed provider, the generation id or cache/reasoning tokens, spend recorded with no run or tenant key, and new model loops with no countable ceiling. Learns the repo's own usage ledger first and holds the new call to it. Use when a PR adds or changes an LLM, embedding or image-generation call.

Ask for it