miiidev/freemaxxing
A local, OpenAI-compatible AI proxy that pools free-tier models across OpenRouter, Groq, Google AI Studio, Mistral, and Cerebras—with smart routing, automatic rate-limit failover, and quota-aware load balancing.
What's novel
A local, OpenAI-compatible AI proxy that pools free-tier models across OpenRouter, Groq, Google AI Studio, Mistral, and Cerebras—with smart routing, automatic rate-limit failover, and quota-aware load balancing.
Code Analysis
14 files read · 4 roundsA local OpenAI-compatible proxy that pools free-tier AI models from multiple providers behind stable aliases, with automatic failover, budget tracking, reliability scoring, and streaming SSE transforms.
Strengths
Genuinely non-trivial implementation: provider-specific failure classification heuristics, multi-criteria model ranking with budget/reliability/speed/tier, streaming SSE transforms for model rewriting and tool-call validation, and a clean PersistedStore abstraction. The quirks system and budget tracking show real depth of understanding of free-tier API limitations.
Weaknesses
The PersistedStore abstraction has some awkwardness — the four `persist*ByStore` functions are no-ops, and the legacy `bind*File` globals coexist with the store, creating dual persistence paths. The SSE transform chain in server.ts is complex and hard to follow, with manual pipe wiring and error handler attachment that could be more declarative.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
117
Contributors
1
Files
70
Active weeks
3
Repository
Language
TypeScript
Stars
1
Forks
0
License
—