[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f1ewo3pm7onkqi":3},{"code":4,"message":5,"data":6,"timestamp":33},10000,"success",{"full_name":7,"position":8,"description":9,"language":10,"stars":11,"forks":12,"stars_gained":13,"age_days":14,"hn_threads":15,"one_liner":21,"summary_md":22,"topic":23,"github_topics":26,"license":27,"homepage":28,"updated_at":29,"history":30},"tashfeenahmed\u002Ffreellmapi",19,"7.4 billion tokens per month. 34 free LLM providers. 635 free model endpoints. All behind one \u002Fv1 endpoint, plus any custom OpenAI-compatible endpoint. Smart routing, automatic failover, encrypted keys. Personal experimentation only.","TypeScript",40804,5504,723,186,[16],{"id":17,"title":18,"points":19,"comments":19,"posted_at":20},49989005,"FreeLLMAPI – One OpenAI-Compatible API for Multiple Free LLM Providers",1,"2026-10-07","One OpenAI-compatible endpoint that pools free tiers from 34 LLM providers with routing and failover, for hobby experiments.","## The problem\n\nNearly every LLM provider hands out a free tier — a few million tokens a month or a few thousand requests a day. Individually that's a toy; stacked, the README claims roughly 7.4B tokens a month. But stacking them by hand means 34 SDKs, 34 rate-limit policies, and 34 failure modes, and quotas change without notice. This project exposes all of the pooled providers behind a single `\u002Fv1` endpoint, routing each request to an available model and failing over when one is throttled.\n\n## What's different\n\nCompared with a gateway like LiteLLM or OpenRouter, the distinguishing idea here is that the product is built around other vendors' *free* quotas: per-key usage tracking to stay under each cap, automatic failover across providers, and encrypted key storage. It also self-updates its model catalog from a signed remote feed, so quota and model changes arrive without a `git pull` — though free installs see a 30-day-old snapshot while the $19\u002Fyr \"premium\" tier gets the same-day feed. Treat that paid gate on freshness as a design choice, not a technical limit.\n\n## Who it's for\n\nDevelopers who want to experiment with coding agents such as Claude Code or Codex CLI, or run personal side projects, without paying per token. Anyone encoding a real product's reliability expectations on top of other companies' free tiers should not bother: the free quotas are not a contract, and the README itself frames the project as personal experimentation only.\n\n## Cost of adoption\n\nTypeScript\u002FNode, MIT licensed, distributed as macOS and Windows desktop builds, a Docker image and a mobile app. The setup cost is not the install but collecting and rotating API keys for each provider — many free tiers require accounts, phone verification or region-bound billing (ModelScope, for instance, reportedly needs an Aliyun binding). The README is long but thin on operational detail for the server side; check the linked architecture docs before self-hosting. Expect provider churn to break routes and the remote catalog feed to be a single point of trust.\n\n## Verdict\n\nA brand-new project: created in April 2026, and with 40.6k stars and one day on the trending board, the interest is concentrated rather than proven over time. The +12k monthly gain is real, but such a fast rise on a project built entirely on other vendors' goodwill invites caution — quotas can shrink and terms can change at any time. It is a clever convenience layer, not infrastructure. Try it for personal agent experiments; don't build a product on it.",{"slug":24,"name":25},"ai-engineering","AI Engineering",[],"MIT","https:\u002F\u002Ffreellmapi.co","2026-10-24T14:37:09Z",[31],{"date":32,"stars":11,"stars_gained":13},"2026-10-24",1792864261]