llmf

Free, keyless LLM access in one call — with automatic multi-provider fallback.

llmf (llmfree) is a stdlib-only Python library and CLI for public, no-API-key OpenAI-compatible endpoints. It picks a live target, falls back on failure, and runs on desktop and Termux.

import llmf

r = llmf.chat("Explain the CAP theorem in two sentences.")
print(r)              # Reply is a str subclass
print(r.source)       # "provider/model"
print(r.provider, r.model, round(r.latency, 3), r.usage)

Install

pip install -U llmf
# optional browser-like TLS fingerprint (still works without it):
pip install -U "llmf[stealth]"

python -c "import llmf; print(llmf.__version__)"
llmf version
llmf -v version          # also prints active transport

pip show llmf

Name: llmf
Version: <installed>     # currently 0.6.0
Summary: free, keyless LLM chat with automatic multi-provider fallback — stdlib only
Home-page: https://github.com/MoonLite-Br/llmfree
Author: Html
License: MIT
Requires-Python: >=3.8
Requires: (none)

Python API

Short form → Reply

import llmf

r = llmf.chat("Hello!")
r.text, r.source, r.provider, r.model, r.latency, r.usage, r.attempts

llmf.chat("hi", model="best")       # aliases: fast | best | code | vision | reasoning
llmf.chat("hi", model=["best", "fast"])
llmf.chat("hi", provider="ch-at")
llmf.chat("hi", provider="nottrack")  # opt-in (auto=False)
llmf.chat("Describe this", images=["photo.jpg"], model="vision")

for piece in llmf.stream("Count 1 to 5"):
    print(piece, end="", flush=True)

OpenAI-shaped form

from llmf import LLMF

llm = LLMF()
c = llm.chat.completions.create(
    model="best",
    messages=[{"role": "user", "content": "Hello"}],
)
print(c.choices[0].message.content)
print(c.model, c.provider, c.usage, c.latency)

Catalog, survival stats, config

for m in llmf.models.list():
    print(m.name, m.tags, m.alive, m.samples, m.streaming)

llmf.models.list(live=True)                 # chat/free by default
llmf.models.list(live=True, include_non_chat=True)
llmf.models.resolve("fast")
llmf.models.check()                         # doctor; appends survival history

llmf.providers.list()
llmf.health.check()

llmf.configure(
    timeout=30,
    vision_timeout=90,
    routing="adaptive",
    race=2,
    transport="auto",          # urllib | auto | curl_cffi
    impersonate="chrome",
)
llmf.register("lab", "http://localhost:11434/v1", ["llama3"], front=True)

Environment: LLMF_TIMEOUT, LLMF_PROXY, LLMF_MODEL, LLMF_ROUTING, LLMF_TRANSPORT, LLMF_IMPERSONATE
Proxy: http://, https://, socks5://, socks5h://, socks4://

What’s new in 0.6.0

• Vetted catalogue — models kept only if they passed ≥2 independent live runs with ≥75% success; weak/sign-in/capacity models removed from the static catalog (uncatalogued provider/model ids still work; use llmf models --live).
• Survival history — each doctor run appends to history.json (last 20 samples per model). llmf models shows alive 7/7, flaky 2/3, dead 0/3. Models with ≥3 samples and <50% success sink to the end of plans. Exposed as ModelInfo.alive / .samples.
• Optional stealth transport — transport="auto"|"curl_cffi", impersonate="chrome", CLI --transport / --impersonate, env LLMF_TRANSPORT / LLMF_IMPERSONATE. Install with pip install "llmf[stealth]" (curl_cffi). Default remains stdlib urllib.
• Streaming metadata — ModelSpec.streams / ModelInfo.streaming; e.g. chatjimmy may show no-stream.
• Tag refresh (fast, vision, reasoning); milder “Retry after N” cooldown (N+1 s).

Full history: CHANGELOG.md.

Built-in providers (auto chain unless noted)

Exact lists change with vetting — always check llmf models / llmf doctor.

CLI

usage: llmf [options] "question"
       llmf chat
       llmf models [list|check|resolve NAME]
       llmf doctor [--all] [--include-non-chat] [--include-limited]
       llmf providers | health
       llmf serve [--port N]
       llmf version

Flags: -m/--model, -p/--provider, -s/--system, --image, --stream, --json, -v, --routing ranked|adaptive, --race N, --timeout, --proxy, --transport urllib|auto|curl_cffi, --impersonate, --live, --all, --include-limited, --include-non-chat, --no-save, serve: --host, --port, --api-key, --cors, --strict-models, --no-fallback.

$ llmf -v "2+2=?"
$ llmf models
$ llmf doctor
$ llmf --transport auto -v version
$ llmf serve --port 8000

Behaviour

• Automatic fallback with cooldowns (rate-limit / capacity / paid differ).
• Survival scores demote chronically failing models in planning.
• Live catalog / doctor default to chat + free only (--include-non-chat to override).
• transport="auto" uses curl_cffi when the extra is installed, else urllib.
• Total failure → llmf.AllProvidersFailed with .errors.

Safety

Unofficial free endpoints may change or vanish. Do not send secrets. Opt-in providers and discover() are explicit. llmf serve defaults to localhost.

Development

git clone https://github.com/MoonLite-Br/llmfree.git
cd llmfree
python -m unittest discover -s tests -v
python -m build && twine check dist/*

License

MIT © 2026 Html / MoonLite-Br