pi-bansos

Free model provider for pi (browse packages). It adds a bansos provider with live free models from 2 upstreams — OpenCode Zen and KiloCode gateway — through a local OpenAI-compatible proxy.

Models (22 total)

All models are free. The provider is one bansos entry, but model names show their upstream: OpenCode or KiloCode. Muse uses the OpenAI Responses API; the other models use Chat Completions. The startup check only verifies catalog membership; upstream access can still change between startup and a request.

OpenCode Zen (8 models)

| Model ID | Name | API | Context | Max Output | Reasoning | |----------|------|-----|---------|------------|-----------| | x-preview-f-free | Ox Alpha Free | chat | 1M tokens | 131K tokens | ✅ | | muse-spark-1.2-contributor-free | Muse Spark 1.2 Free | responses | 1M tokens | 131K tokens | ✅ | | mimo-v2.5-free | MiMo V2.5 Free | chat | 200K tokens | 32K tokens | ✅ | | hy3-free | Hy3 Free | chat | 190K tokens | 64K tokens | ✅ | | nemotron-3-ultra-free | Nemotron 3 Ultra Free | chat | 1M tokens | 128K tokens | ✅ | | nemotron-3.5-lightning-free | Nemotron 3.5 Lightning Free | chat | 262K tokens | 262K tokens | ✅ | | big-pickle | Big Pickle | chat | 200K tokens | 32K tokens | ✅ | | laguna-s-2.1-free | Laguna S 2.1 Free | chat | 256K tokens | 32K tokens | ✅ |

Muse note: Muse uses OpenAI Responses (/v1/responses), while the other OpenCode models use Chat Completions (/v1/chat/completions). pi-bansos selects the API per model and suppresses Muse's unsupported reasoning.effort: "none" value when reasoning is off. It was verified end-to-end through both the proxy and pi-ai adapter with a response of OK.

Manual check: select OpenCode · Muse Spark 1.2 Free in /model, then ask it to Reply with exactly OK.

Removed after direct inference checks: deepseek-v4-flash-free (free promotion ended), north-mini-code-free, and ling-3.0-flash-free (not supported).

KiloCode Gateway (14 models)

Keyless — 200 requests/hour per IP.

| Model ID | Name | Context | Max Output | Reasoning | |----------|------|---------|------------|-----------| | kilo-auto/free | Kilo Auto Free | 256K tokens | 10K tokens | ❌ | | stepfun/step-3.7-flash:free | Step 3.7 Flash Free | 262K tokens | 262K tokens | ❌ | | nvidia/nemotron-3-ultra-550b-a55b:free | Nemotron 3 Ultra Free | 1M tokens | 65K tokens | ✅ | | nvidia/nemotron-3-super-120b-a12b:free | Nemotron 3 Super Free | 262K tokens | 262K tokens | ✅ | | dots-studio/dots-3-note-preview:free | Dots3-Note Preview Free | 512K tokens | 512K tokens | ✅ | | cohere/north-mini-code:free | North Mini Code Free | 256K tokens | 64K tokens | ❌ | | poolside/laguna-xs-2.1:free | Laguna XS 2.1 Free | 262K tokens | 32K tokens | ❌ | | nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free | Nemotron 3 Nano Omni Free | 256K tokens | 65K tokens | ✅ | | openrouter/free | OpenRouter Free (auto) | 200K tokens | 65K tokens | ❌ | | nvidia/nemotron-3.5-lightning:free | Nemotron 3.5 Lightning Free | 1M tokens | 65K tokens | ✅ | | nvidia/nemotron-3.5-content-safety:free | Nemotron 3.5 Content Safety Free | 128K tokens | 8K tokens | ✅ | | tencent/hy3:free | Tencent Hy3 Free | 262K tokens | 128K tokens | ✅ | | liquid/lfm-2.5-2.6b:free | Liquid LFM 2.5 2.6B Free | 128K tokens | 8K tokens | ❌ | | poolside/laguna-s-2.1:free | Laguna S 2.1 Free | 262K tokens | 32K tokens | ✅ |

Rate limiting is separated internally by upstream: OpenCode uses its UTC-day local guard and its own upstream free quota; Kilo uses a rolling-hour local guard matching its documented 200/hour/IP limit.

Why pi-bansos

• Zero cost — all models free, no API key needed for supported upstreams
• Auto health-check — only catalog-listed models registered at startup; dead ones skipped silently
• 22 models from 2 sources — OpenCode Zen + KiloCode gateway
• Local-only proxy — binds to 127.0.0.1, nothing exposed externally
• Optional relay egress — route through a Vercel/Cloudflare relay to dodge per-IP rate limits, toggled live via /bansos
• Auto port bump — if port 18080 is taken, automatically tries the next one (up to 18100)

Install

Requires pi.

pi install npm:pi-bansos

Usage

pi
# /model → bansos → choose a free model

Run /bansos any time to toggle relay egress or switch between saved relays (see Relay).

Optional custom port:

BANSOS_PORT=18081 pi

Relay (optional)

By default pi-bansos talks to the free upstreams directly. If your IP gets rate-limited or blocked, switch on a relay — requests then go out through a relay worker instead of your own IP. Toggle it live from inside pi, no restart:

| Command | What it does | | --- | --- | | /bansos on | Route through the relay | | /bansos off | Go direct (default) | | /bansos status | Show current state, request count, and saved-relay count | | /bansos url <URL> | Use a different relay (added to saved list) | | /bansos use <URL> | Switch to a relay and enable it (added to saved list) | | /bansos list | Show all saved relays (★ = active) | | /bansos remove <URL> | Forget a saved relay (the active one can't be removed) | | /bansos deploy | Deploy a fresh Vercel relay and switch to it | | /bansos | Interactive menu (incl. Switch / Remove relay…) |

The state is saved at the package root (.relay-state.json, next to the extensions/ folder) and remembered across restarts — you manage it only via /bansos, nothing in your shell. Every relay you deploy, use, or url is kept in a saved list, so you can switch between them anytime without re-typing URLs. Any HTTP relay works (Vercel, Cloudflare, Deno, or your own). There is no built-in default — run /bansos deploy to create one or /bansos url <URL> to use your own.

Switching between saved relays (e.g. you deployed one and also have another):

/bansos list
  Saved relays (2):
  ★ https://pi-bansos-relay-xxxx.vercel.app   [deployed relay-2026]
    https://vercel-relay-yyyy.vercel.app       [manual]

/bansos            → Switch relay… → pick one → active (live, no restart)
/bansos use https://vercel-relay-yyyy.vercel.app   # or switch directly

/bansos deploy — one-command Vercel relay

Deploys your own Node.js relay to Vercel and activates it immediately. It asks for a Vercel API token (get one at https://vercel.com/account/tokens) and an optional project name, then uploads a tiny worker and waits for it to go live (~10–40 s). The new relay URL is saved and switched on; the token is used once and never stored.

/bansos deploy
  Vercel API token (vercel-…): <paste>
  Project name (empty = auto):  relay-2026
  Uploading relay to Vercel…
  Waiting for deployment to go live…
  ✓ Deployed & active: https://relay-2026-xxx.vercel.app

The Vercel relay masks your IP behind Vercel's dynamic edge IPs. Free tier: 100 GB bandwidth + 500 K invocations/month. Deploy on multiple accounts for more IP diversity. The token input has no hidden/secret mode in the TUI, so it shows while typing — paste, deploy, done.

A relay is a single fixed exit IP, not rotation. Useful when your IP is limited; otherwise it just adds a small hop.

Notes

• Free upstream models are best-effort: promos can expire, model IDs can change, and rate limits may apply
• pi-bansos health-checks at startup so unavailable models are skipped instead of registered
• KiloCode gateway: 200 req/hr per IP, keyless

Uninstall

pi remove npm:pi-bansos

License

MIT