diff --git a/CHANGELOG.md b/CHANGELOG.md index ec69566..cb4c61d 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -1,5 +1,13 @@ # Changelog +## 0.2.11 — 2026-09-12 + +### Changed + +- **RESOURCES.md** is now a step-by-step **Free Student Ai Tokens** guide (use-it-or-lose-them, Workers Ai first, wired hops, Tier-2 mint names, key safety). +- Exported optional laptop path: [docs/free-llm-api.md](docs/free-llm-api.md) (FreeLLMAPI attribution + student steps). +- README and live `/resources` point at the new guide. + ## 0.2.10 — 2026-09-10 ### Changed diff --git a/README.md b/README.md index 7f87a89..a837993 100644 --- a/README.md +++ b/README.md @@ -74,7 +74,7 @@ Season dates / brackets: still being posted on [ROADMAP.md](ROADMAP.md). The tab | You want | You need | Skip | |---|---|---| -| Free Ai keys only (no tournament) | [RESOURCES.md](RESOURCES.md) | Omarchy, Larga, CloudFlare | +| Free Student Ai tokens (step-by-step, use-it-or-lose-them) | [RESOURCES.md](RESOURCES.md) · optional laptop `/v1`: [docs/free-llm-api.md](docs/free-llm-api.md) | Omarchy until you want the Cup | | A Linux laptop built for this | [Omarchy](https://omarchy.org/) | Larga until you want the Cup | | **Enter the Cup** / Board / Tape | Open the live URL **on that phone** | extra keys until Drive 429s | | Android terminal | **Termux from [F-Droid](https://f-droid.org/packages/com.termux/)** | Play Store Termux | @@ -148,7 +148,7 @@ npx wrangler d1 migrations apply larga npm run deploy ``` -When Drive starts 429ing, add keys from [RESOURCES.md](RESOURCES.md) with **`wrangler secret put` only**: +**Free tokens:** [RESOURCES.md](RESOURCES.md) is the step-by-step guide — Step 0 = shared Cup with **0 keys** (Workers Ai); Step 2 = wired hops via **`wrangler secret put` only**: ```sh npx wrangler secret put GROQ_API_KEY @@ -156,9 +156,11 @@ npx wrangler secret put OPENROUTER_API_KEY npx wrangler secret put GEMINI_API_KEY ``` -**Safe:** the secret stays on CloudFlare. **Not safe:** pasting a key into Grid, GitHub, Discord, a screenshot, or any file you `git add`. If it leaked, revoke it and mint a new one. +Quotas nap and reset — **use them or lose them.** Drive hops on 429/5xx/529/empty; **401 stops the hop** so a dead key does not burn the next mint. -More catalogs (verify dates, don’t fork): [Student-free-ai-packs](https://github.com/asbinthapa99/Student-free-ai-packs-) · [FreeLLMAPI](https://github.com/tashfeenahmed/freellmapi) (optional laptop `/v1`, not the Cup). +**Safe:** secrets stay on CloudFlare. **Not safe:** pasting a key into Grid, GitHub, Discord, a screenshot, or any file you `git add`. Leaked? Revoke and mint new. + +Optional laptop router (not the Cup): [docs/free-llm-api.md](docs/free-llm-api.md) · upstream [FreeLLMAPI](https://github.com/tashfeenahmed/freellmapi). Catalogs (verify dates): [Student-free-ai-packs](https://github.com/asbinthapa99/Student-free-ai-packs-). GitHub has **Copy** on code blocks. Live `/install` and `/resources` have **Copy** too. diff --git a/RESOURCES.md b/RESOURCES.md index adad7e1..2232ef9 100644 --- a/RESOURCES.md +++ b/RESOURCES.md @@ -1,38 +1,55 @@ -# Free Student Ai +# Free Student Ai Tokens — use them or lose them -Skip Larga. Skip Omarchy. This page is the tokens. +**The viral sell:** every vendor gives you free Ai quota. It **resets on a schedule** — daily Neurons, monthly caps, RPM windows. **Use it today or it vanishes.** When one mint naps, hop to the next. When a key dies with **401**, Drive **stops** — so a dead key does not burn the next mint. -Play the shared Cup at [larga.openroyleal.com](https://larga.openroyleal.com) with **no keys**. Workers Ai is already on. +Play the shared Cup at [larga.openroyleal.com](https://larga.openroyleal.com) with **zero keys**. CloudFlare Workers Ai is already on the kettle. -Quotas die. Make three accounts. When one naps, use the next. +Quotas die. Make three accounts. When one naps, use the next. Revoke + mint if a key leaked. -## Keep keys safe +--- -A token is a password for Ai. +## Step 0 — Play the Cup with 0 keys (Workers Ai) -| Do | Don’t | -|---|---| -| `npx wrangler secret put GROQ_API_KEY` on **your** Worker | Paste keys into Grid, Discord, GitHub, a screenshot, or `.dev.vars` you might commit | -| Play the shared Cup first (our kettle, no student keys) | Put your Groq/Gemini key on the shared Cup — there is nowhere to paste it, on purpose | -| Revoke + mint new if it leaked | Share a key with a classmate “just this once” | +**Unli rice.** CloudFlare Workers Ai ships with every Larga deploy — including the shared Cup. No Groq. No Gemini paste box. Just Grid. -`.dev.vars` is gitignored and is for **local fake secrets** (`CERT_SECRET`), not vendor API keys you care about. +1. Open [larga.openroyleal.com](https://larga.openroyleal.com) +2. Continue with GitHub or Google +3. Grid → run a chain -## Catalogs (trust but verify) +**Drive** tries Workers Ai first (`@cf/meta/llama-3.2-1b-instruct`, then `@cf/meta/llama-3.1-8b-instruct-fast`). Free Neurons reset on CloudFlare’s schedule — **use them or lose them.** When Workers Ai 429s or empties, Drive hops (if you added keys in Step 2). -We do **not** fork these. We link them. Deals rot. +The shared Cup has **nowhere to paste your vendor keys** — on purpose. Your mints stay on **your** Worker. -| Catalog | What it is | Use for | -|---|---|---| -| [Student-free-ai-packs](https://github.com/asbinthapa99/Student-free-ai-packs-) | Student software / cloud / Pack deals (~5 months old as of check) | GitHub Student Pack, JetBrains, Figma Education, Azure for Students, AWS Educate. **Skip expired rows** (e.g. Google One Gemini Advanced “before June 30 2025”). Heroku “free tier” is gone. | -| [GitHub Student Pack](https://education.github.com/pack) | The live Pack | Copilot / Codespaces / whatever GitHub still lists today | -| [FreeLLMAPI](https://github.com/tashfeenahmed/freellmapi) | Self-host one `/v1` that hops many free mints | Optional. Not required for Larga. Personal use. Live model list: [freellmapi.co/models](https://freellmapi.co/models.html) | +--- + +## Step 1 — One CloudFlare account + `npm run deploy` + +Your own Worker = your own Board + your own kettle secrets. + +**Only from repo root.** `packages/` is source. Do **not** bind `larga.openroyleal.com` in your `wrangler.jsonc` — that hostname is the shared Cup (`npm run deploy:cup` is Agent OpenRoyleAl only). + +[![Deploy to Cloudflare](https://deploy.workers.cloudflare.com/button)](https://deploy.workers.cloudflare.com/?url=https://github.com/OpenRoyleAl/Larga) + +```sh +git clone https://github.com/OpenRoyleAl/Larga +cd Larga +npm install +npx wrangler login +npx wrangler d1 create larga +# paste database_id into wrangler.jsonc (repo root only) +npx wrangler d1 migrations apply larga +npm run deploy +``` + +Workers Ai is already wired (`"ai": { "binding": "AI" }` in root `wrangler.jsonc`). **No extra key** for the default kettle. -Larga Drive is our hop on CloudFlare. FreeLLMAPI is if you want a laptop `/v1` without deploying Larga. +Smoke: `/health`, `/install`, `/grid`, `/resources` on your `*.workers.dev` URL. -## After you deploy Larga (`npm run deploy` from repo root) +--- -Workers Ai is already on. Extra keys keep the Cup alive: +## Step 2 — Wired hops (mint now, Drive uses them today) + +When Workers Ai naps, add vendor keys with **`npx wrangler secret put` on your Worker only**: ```sh npx wrangler secret put GROQ_API_KEY @@ -40,28 +57,92 @@ npx wrangler secret put OPENROUTER_API_KEY npx wrangler secret put GEMINI_API_KEY ``` -Drive hops on 429 / 5xx / 529 / empty / quota text. A **401** stops the hop so a dead key does not burn the next mint. +| Secret | Mint where | Drive hop order (after Workers Ai) | +|---|---|---| +| `GROQ_API_KEY` | [console.groq.com](https://console.groq.com/) | Groq `llama-3.1-8b-instant` | +| `OPENROUTER_API_KEY` | [openrouter.ai](https://openrouter.ai/) | OpenRouter `meta-llama/llama-3.1-8b-instruct:free` | +| `GEMINI_API_KEY` | [aistudio.google.com/apikey](https://aistudio.google.com/apikey) | Gemini `gemini-2.0-flash` | + +**Hop rules (Drive):** 429, 5xx, 529, empty body, quota text → try the next provider. **401 / 403 / 404 → stop.** Fix or revoke the bad key before the next mint runs. + +Optional (same pattern, not required for Cup): `OPENAI_API_KEY` if you have paid or trial OpenAI credits. + +Gmail works for Groq, Ai Studio, and OpenRouter. School email helps elsewhere (Step 5). + +--- + +## Step 3 — More free mints (mint now; Drive wiring coming) + +Stack free tiers **before** Drive knows about them — mint the keys, hold them safe, add to your Worker when we wire the hop. + +| Mint | Where to mint | Planned `wrangler secret put` name | Drive today? | +|---|---|---|---| +| Hugging Face | [huggingface.co/settings/tokens](https://huggingface.co/settings/tokens) | `HUGGINGFACE_API_KEY` | mint only | +| Cerebras | [cloud.cerebras.ai](https://cloud.cerebras.ai/) | `CEREBRAS_API_KEY` | mint only | +| SambaNova | [cloud.sambanova.ai](https://cloud.sambanova.ai/) | `SAMBANOVA_API_KEY` | mint only | +| Mistral | [console.mistral.ai](https://console.mistral.ai/) | `MISTRAL_API_KEY` | mint only | +| GitHub Models | [github.com/marketplace/models](https://github.com/marketplace/models) | `GITHUB_MODELS_TOKEN` | mint only | + +**Use it or lose them:** each vendor meters RPM/RPD/TPM/TPD. Free windows reset at UTC midnight or on a rolling clock — check the vendor dashboard, not this doc. When one account naps, use the next account you minted. + +When Drive adds a hop, the secret name above is what you will run — same `wrangler secret put` rule as Step 2. + +--- + +## Step 4 — FreeLLMAPI path (optional laptop `/v1`) -## Mints we wire or list +**Primary path stays Larga Drive on CloudFlare** (Steps 0–2). **FreeLLMAPI** is optional: one local OpenAI-compatible `/v1` that routes across many free mints on your laptop — useful for Codex, Claude Code, Cursor, or scripts **off** the Cup. -| Mint | Link | Drive? | +We exported the student how-to into the repo so you can follow offline: + +→ **[docs/free-llm-api.md](docs/free-llm-api.md)** — install, keys, `auto` routing, trust-but-verify, link back to [upstream FreeLLMAPI](https://github.com/tashfeenahmed/freellmapi). + +Live model catalog (deals rot): [freellmapi.co/models](https://freellmapi.co/models.html). + +--- + +## Step 5 — Student Pack / Azure / AWS / JetBrains / Figma + +Software and cloud **credits** — not Drive hops. Still **use it or lose them** (Pack renewals, credit expiry). + +| Deal | Link | Good for | |---|---|---| -| CloudFlare Workers Ai | with Larga deploy | yes, default | -| Groq | [console.groq.com](https://console.groq.com/) | `GROQ_API_KEY` | -| Google Ai Studio | [aistudio.google.com/apikey](https://aistudio.google.com/apikey) | `GEMINI_API_KEY` | -| OpenRouter `:free` | [openrouter.ai](https://openrouter.ai/) | `OPENROUTER_API_KEY` | -| Hugging Face | [huggingface.co/settings/tokens](https://huggingface.co/settings/tokens) | mint only | -| Cerebras | [cloud.cerebras.ai](https://cloud.cerebras.ai/) | mint only | -| SambaNova | [cloud.sambanova.ai](https://cloud.sambanova.ai/) | mint only | -| Mistral | [console.mistral.ai](https://console.mistral.ai/) | mint only | -| GitHub Models | [github.com/marketplace/models](https://github.com/marketplace/models) | mint only | -| GitHub Student Pack | [education.github.com/pack](https://education.github.com/pack) | Copilot / credits, not Drive | +| GitHub Student Pack | [education.github.com/pack](https://education.github.com/pack) | Copilot, Codespaces, whatever GitHub lists today | | JetBrains Edu | [jetbrains.com/community/education](https://www.jetbrains.com/community/education/) | IDEs | | Figma Education | [figma.com/education](https://www.figma.com/education/) | design | | Azure for Students | [azure.microsoft.com/free/students](https://azure.microsoft.com/en-us/free/students/) | cloud credits | | AWS Educate | [aws.amazon.com/education/awseducate](https://aws.amazon.com/education/awseducate/) | cloud credits | -School email helps Pack / Azure / JetBrains. Gmail still works on Groq + Ai Studio + OpenRouter. +Verify dates yourself. We link catalogs; we do not fork stale deal lists. + +--- + +## Keep keys safe + +A token is a password for Ai. + +| Do | Don’t | +|---|---| +| `npx wrangler secret put GROQ_API_KEY` on **your** Worker | Paste keys into Grid prompts, Discord, GitHub, a screenshot, or `.dev.vars` you might commit | +| Play the shared Cup first (our kettle, no student keys) | Put your Groq/Gemini key on the shared Cup — there is nowhere to paste it, on purpose | +| Revoke + mint new if it leaked | Share a key with a classmate “just this once” | +| Hop on 429/5xx/529/empty; fix 401 before the next mint | Assume unlimited forever — quotas die | + +`.dev.vars` is gitignored and is for **local fake secrets** (`CERT_SECRET`, `SERVICE_SECRET`), not vendor API keys you care about. + +--- + +## Catalogs (trust but verify) + +We **do not** fork these. We link them. Deals rot. + +| Catalog | What it is | Use for | +|---|---|---| +| [Student-free-ai-packs](https://github.com/asbinthapa99/Student-free-ai-packs-) | Student software / cloud / Pack deals | Cross-check Step 5; **skip expired rows** | +| [GitHub Student Pack](https://education.github.com/pack) | The live Pack | Copilot / credits | +| [FreeLLMAPI upstream](https://github.com/tashfeenahmed/freellmapi) | Self-host one `/v1` that hops many free mints | Optional laptop path — see [docs/free-llm-api.md](docs/free-llm-api.md) | + +--- ## Android terminal @@ -72,6 +153,23 @@ pkg update && pkg upgrade pkg install tmux git ``` +Deploy Larga from Termux the same way as Step 1, or run FreeLLMAPI locally — see [docs/free-llm-api.md](docs/free-llm-api.md). + ## iPhone Play in Safari. Share → Add to Home Screen. There is no Termux. You do not need a terminal. + +--- + +## Quick reference — mint → secret → Drive + +| Kettle | Keys needed on shared Cup | Secret on your Worker | +|---|---|---| +| CloudFlare Workers Ai | **0** | already bound at deploy | +| Groq | mint yourself | `GROQ_API_KEY` ✓ wired | +| OpenRouter `:free` | mint yourself | `OPENROUTER_API_KEY` ✓ wired | +| Google Ai Studio | mint yourself | `GEMINI_API_KEY` ✓ wired | +| HF / Cerebras / SambaNova / Mistral / GitHub Models | mint yourself | names in Step 3 — mint only until Drive wires them | +| FreeLLMAPI | N/A (local `/v1`) | see [docs/free-llm-api.md](docs/free-llm-api.md) | + +*Larga na.* Born in Cebu. diff --git a/docs/free-llm-api.md b/docs/free-llm-api.md new file mode 100644 index 0000000..71406c5 --- /dev/null +++ b/docs/free-llm-api.md @@ -0,0 +1,171 @@ +# FreeLLMAPI — optional laptop `/v1` (exported guide) + +> **Attribution:** This doc extracts the **student-actionable** parts of [FreeLLMAPI](https://github.com/tashfeenahmed/freellmapi) by [tashfeenahmed](https://github.com/tashfeenahmed). Upstream is MIT-licensed. We do **not** vendor the FreeLLMAPI runtime inside Larga — install it separately when you want a local router. +> +> **Trust but verify:** Free tiers, model lists, and quotas **rot weekly**. Numbers below came from upstream docs and [freellmapi.co/models](https://freellmapi.co/models.html) — check upstream before you depend on them. +> +> **Larga primary path:** [RESOURCES.md](../RESOURCES.md) Steps 0–2 — CloudFlare Workers Ai + Drive on **your** Worker. FreeLLMAPI is for **optional** laptop tooling (Codex, Claude Code, Cursor, scripts), not the Cup Board. + +--- + +## What FreeLLMAPI is + +One **OpenAI-compatible** endpoint (`http://localhost:3001/v1`) that: + +- Stores **your** vendor keys encrypted (AES-256-GCM in local SQLite) +- Routes each request to the best available **free** model +- **Fails over** on 429/5xx to the next model in your chain +- Tracks per-key RPM/RPD/TPM/TPD so you stay under caps +- Gives you one **unified** bearer token (`freellmapi-…`) for all clients + +Upstream stacks many free tiers (Groq, Google, Cerebras, OpenRouter, CloudFlare, Hugging Face, Mistral, …). The live catalog lists hundreds of free model endpoints — browse at **[freellmapi.co/models](https://freellmapi.co/models.html)**. + +**Use it or lose them:** each provider’s free quota resets on its own clock. The router hops when one mint naps — same spirit as Larga Drive, but on your machine. + +--- + +## When to use FreeLLMAPI vs Larga Drive + +| You want | Use | +|---|---| +| Enter the Cup, Board, Tape on a phone | [larga.openroyleal.com](https://larga.openroyleal.com) or your Larga deploy — **0 keys** on shared Cup | +| Hop on CloudFlare without a laptop server | Larga Drive (`npm run deploy`) — Workers Ai + optional `wrangler secret put` | +| Point Codex / Claude Code / Cursor at one local `/v1` | FreeLLMAPI on laptop (this doc) | +| Keep keys off Discord/GitHub/Grid | Both: Larga uses `wrangler secret put`; FreeLLMAPI uses its **Keys** dashboard — never paste vendor keys into chat | + +--- + +## Quick start (thin path — Docker one-liner) + +**Prerequisites:** Docker installed. + +```bash +curl -fsSL https://freellmapi.co/install.sh | bash +``` + +Re-running is safe — your `.env` and encryption key are preserved. Read the script first if you prefer: [freellmapi.co/install.sh](https://freellmapi.co/install.sh). + +1. Open **http://localhost:3001** +2. **Keys** page → add provider keys (Groq, Google, OpenRouter, CloudFlare, … — same mints as [RESOURCES.md](../RESOURCES.md)) +3. Reorder the **Fallback Chain** +4. Copy the **unified API key** from the Keys header + +**Desktop (no Docker):** [Releases `.exe` / `.dmg`](https://github.com/tashfeenahmed/freellmapi/releases/latest) — tray app, no password on desktop builds. + +Full install matrix (Compose, dev, LAN, Android Termux): [upstream install doc](https://github.com/tashfeenahmed/freellmapi/blob/main/docs/en/install/01-install.md). + +--- + +## Add your free mints (Keys page) + +Mint keys at the same places as Larga RESOURCES Step 2–3: + +| Platform | Mint link | +|---|---| +| Groq | [console.groq.com](https://console.groq.com/) | +| Google Ai Studio | [aistudio.google.com/apikey](https://aistudio.google.com/apikey) | +| OpenRouter | [openrouter.ai](https://openrouter.ai/) | +| CloudFlare | Workers Ai / API token per upstream dashboard | +| Hugging Face | [huggingface.co/settings/tokens](https://huggingface.co/settings/tokens) | +| Cerebras | [cloud.cerebras.ai](https://cloud.cerebras.ai/) | +| Mistral | [console.mistral.ai](https://console.mistral.ai/) | + +Paste each key into FreeLLMAPI’s **Keys** UI — **not** into Grid, GitHub, or Discord. FreeLLMAPI encrypts them at rest; your apps only see the unified `freellmapi-…` token. + +**Leaked a vendor key?** Revoke at the vendor, mint new, update Keys page. Same rule as Larga. + +--- + +## Call the API (`auto` routing) + +Point any OpenAI SDK at your local base URL: + +```python +from openai import OpenAI + +client = OpenAI( + base_url="http://localhost:3001/v1", + api_key="freellmapi-your-unified-key", +) + +resp = client.chat.completions.create( + model="auto", # or "auto:fast", "auto:smart", a profile, or a model id + messages=[{"role": "user", "content": "Summarise the fall of Rome in one sentence."}], +) +print(resp.choices[0].message.content) +print("Routed via:", resp.headers.get("x-routed-via")) +``` + +**Routing idea:** `auto` lets the router pick the best healthy model under quota. Named models pin to one entry; profiles (e.g. coding chain) switch fallback order. Responses include `X-Routed-Via` so you see which provider actually served. + +curl, streaming, tools, vision, embeddings: [upstream API reference](https://github.com/tashfeenahmed/freellmapi/blob/main/docs/en/api/01-rest-api.md). + +--- + +## Coding agents (one command setup) + +With FreeLLMAPI running locally: + +```bash +npx freellmapi setup-claude --url http://localhost:3001 --api-key +``` + +Other generators: `setup-codex`, `setup-aider`, `setup-cursor` (guide), `setup-continue`, … — full list in [upstream clients doc](https://github.com/tashfeenahmed/freellmapi/blob/main/docs/en/clients/01-agent-clients.md). + +| Agent | Setup command | Base URL | +|---|---|---| +| Claude Code | `setup-claude` | root | +| Codex CLI | `setup-codex` | `/v1` | +| Continue / Aider / Cline | `setup-continue` / `setup-aider` / `setup-cline` | `/v1` | + +Launchers (`freellmapi launch`, `launch-codex`) inject credentials without writing them to config files. + +--- + +## Model catalog (live feed) + +Upstream maintains a **signed catalog** synced twice daily — new free models, quota changes, compatibility fixes. + +- **Browse live:** [freellmapi.co/models](https://freellmapi.co/models.html) +- **Free installs** pull a monthly snapshot (~30 days behind live feed) +- **Premium** ($19/yr upstream) gets same-day catalog updates — optional; Larga docs do not require it + +Example provider families in the catalog (verify live): NavyAI, HuggingFace Router, CloudFlare Workers AI, Cohere, AI Horde, Groq, Google, Cerebras, Mistral, OpenRouter, … + +**Do not treat listed “7.4B tokens/month” as unlimited forever** — it is the sum of **metered free tiers**, each with its own reset and nap behavior. + +--- + +## Limitations (honest) + +From upstream — still true for students: + +- No frontier models; latency varies; no SLA +- Top models hit daily caps late in the UTC day, then reset +- Free tiers are for **learning and prototyping**, not production +- Comply with each vendor’s Terms of Service when traffic is proxied + +Full upstream disclaimer: [FreeLLMAPI README § Disclaimer](https://github.com/tashfeenahmed/freellmapi#disclaimer). + +--- + +## Uninstall / data location + +| OS | Data folder | +|---|---| +| Windows | `%APPDATA%\FreeLLMAPI\` | +| macOS | `~/Library/Application Support/FreeLLMAPI/` | +| Linux | `~/.config/FreeLLMAPI/` | + +Docker: `docker compose down -v` drops the volume. Uninstalling the app does **not** delete keys until you remove the data folder. + +--- + +## Link back + +- **Repo:** [github.com/tashfeenahmed/freellmapi](https://github.com/tashfeenahmed/freellmapi) +- **Site:** [freellmapi.co](https://freellmapi.co/) +- **Models:** [freellmapi.co/models](https://freellmapi.co/models.html) +- **Larga Cup path:** [RESOURCES.md](../RESOURCES.md) + +If this exported guide drifts from upstream, trust the repo and site — open an issue on Larga if our steps are wrong, or PR upstream if FreeLLMAPI itself changed. diff --git a/packages/dex/src/pages.ts b/packages/dex/src/pages.ts index 2dee4eb..65e5a88 100644 --- a/packages/dex/src/pages.ts +++ b/packages/dex/src/pages.ts @@ -164,20 +164,23 @@ export function resourcesPage(): string {

Larga

Free Student Ai -

Play the Cup at larga.openroyleal.com with no keys. Skip Larga if you only want tokens. Live list: repo RESOURCES.md. Extra catalogs (verify): Student packs · FreeLLMAPI.

+

Use it or lose them. Free quotas nap and reset. Step 0: play the Cup at larga.openroyleal.com with 0 keys — Workers Ai is already on. Full guide: repo RESOURCES.md · optional laptop FreeLLMAPI.

+
+

Step 0 — 0 keys

+

Shared Cup = CloudFlare Workers Ai kettle. No paste box for your vendor keys — on purpose.

+

Keep keys safe

A token is a password for Ai. Treat it like one.

-

Play on the shared Cup first — you do not paste keys there. Own Worker: npx wrangler secret put … only. Never git, Discord, Grid prompts, or screenshots. Leaked? Revoke it on the vendor site and mint a new one.

+

Your Worker only: npx wrangler secret put …. Never git, Discord, Grid prompts, screenshots, or .dev.vars for real vendor keys. Leaked? Revoke + mint. 401 stops Drive hop — fix the dead key before the next mint.

-

Groq · Google Ai Studio · OpenRouter · Hugging Face

-

GitHub Student Pack · Cerebras · SambaNova · Mistral · GitHub Models

-

CloudFlare Workers Ai ships with a Larga deploy (no extra key).

+

Step 2 — wired hops

+ ${snip("secrets", "npx wrangler secret put GROQ_API_KEY\nnpx wrangler secret put OPENROUTER_API_KEY\nnpx wrangler secret put GEMINI_API_KEY")} +

Step 3 mints (HF, Cerebras, SambaNova, Mistral, GitHub Models): see RESOURCES.md — mint only until Drive wires them.

-

Keep Larga Drive alive

- ${snip("secrets", "npx wrangler secret put GROQ_API_KEY\nnpx wrangler secret put OPENROUTER_API_KEY\nnpx wrangler secret put GEMINI_API_KEY")} +

Groq · Google Ai Studio · OpenRouter · Hugging Face · Student Pack

Android