infreerence

KO | EN

gitlite — search

#claude #containers #git #editor #codex #python #web #awesome-list #react-native #kotlin

(★ 9)

Find exposed AI inference endpoints

파일 탐색기

ZIP 다운로드

- infreerence.png
- - __init__.py
  - chat_routes.py
  - chats_routes.py
  - integrations_routes.py
  - keys_routes.py
  - loot_routes.py
  - models.py
  - overview_routes.py
  - routes.py
  - scans_routes.py
  - verify_routes.py
- - __init__.py
  - store.py
- - __init__.py
  - base.py
  - registry.py
  - shodan.py
- - __init__.py
  - manager.py
  - store.py
- - __init__.py
  - schema.py
  - store.py
- - __init__.py
  - base.py
  - litellm.py
  - llamacpp.py
  - ollama.py
  - openai.py
  - openrouter.py
  - registry.py
  - webservice.py
- - app.js
  - index.html
  - styles.css
- __init__.py
- __main__.py
- app.py
- banner.py
- chat.py
- cli.py
- config.py
- exchange.py
- integrations.py
- keys.py
- loot.py
- overview.py
- ranking.py
- scope.py
- services.py
- term.py
- verification.py
- infreerence_entry.py
- - gpustack.yaml
  - koboldcpp.yaml
  - litellm-proxy.yaml
  - litellm-web.yaml
  - llamacpp-server.yaml
  - llamacpp-web.yaml
  - lm-studio.yaml
  - localai.yaml
  - lorax.yaml
  - ollama-running.yaml
  - one-api.yaml
  - open-webui.yaml
  - openrouter-gateway.yaml
  - sglang.yaml
  - vllm-web.yaml
  - vllm.yaml
  - xinference.yaml
- test_api.py
- test_banner.py
- test_chat.py
- test_chats.py
- test_cli.py
- test_exchange.py
- test_integrations.py
- test_keys.py
- test_litellm_verifier.py
- test_llamacpp_verifier.py
- test_loot.py
- test_ollama_verifier.py
- test_openai_verifier.py
- test_openrouter_verifier.py
- test_overview.py
- test_ranking.py
- test_scans_store.py
- test_scope.py
- test_services.py
- test_shodan_provider.py
- test_templates.py
- test_term.py
- test_verifiers.py
- test_webservice_verifier.py
- API-Keys.md
- API-Reference.md
- Concepts.md
- Development.md
- Home.md
- Integrations.md
- Model-Verification.md
- Scanning.md
- Stop-and-Resume.md
- Throughput.md
- Verify-Hosts.md
- Web-UI.md
.env.example
.gitignore
infreerence.spec
install.ps1
install.sh
LICENSE
pyproject.toml
README.md
uv.lock

# CDN으로 사용하기

jsDelivr

jsDelivr는 공개 GitHub 리포지토리를 별도 설정 없이 CDN으로 즉시 서빙합니다. 버전과 파일을 고르면 웹페이지에 바로 붙일 수 있는 링크와 예시 코드가 만들어집니다.

버전

파일

링크

예시

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<h1 align="center">infreerence</h1>

Find exposed AI inference endpoints.

Recon engine for internet-exposed AI infrastructure. Curated YAML templates run
 search queries against recon sources, then verifiers probe every hit to answer the
 only question that matters: does it actually serve inference?
 It also finds leaked provider API keys in repos and folders, and validates them.

<a href="wiki/Home.md">📖 Documentation</a> ·
 <a href="wiki/Scanning.md">Scanning</a> ·
 <a href="wiki/API-Keys.md">API keys</a> ·
 <a href="wiki/API-Reference.md">HTTP API</a>

---

## ⚠️ Legal & authorized use

By default this tool **sends real HTTP requests to live hosts**, including chat
completions against the inference endpoints it finds (which means it burns someone
else's tokens), and the author is **not responsible for any misuse of it**.
Use it **only against your own infrastructure, or environments you are explicitly
authorized to test**.

---

## Install

```bash
./install.sh          # installs infreerence + the LiteLLM bridge, both on your PATH
```

Uses [`uv`](https://docs.astral.sh/uv/) if present, falls back to a pip venv. Or do it yourself:

```bash
uv tool install .                    # just the CLI, isolated
pip install '.[bridge]'              # or: CLI + bridge in one venv (Python ≥ 3.10)
```

Then:

```bash
cp .env.example .env                 # add your SHODAN_API_KEY
infreerence init                     # set up ~/.infreerence — run once
infreerence --help
```

## Quick start

```bash
infreerence scan ollama # search + verify a provider, live progress bar
infreerence show <scan_id> # what it found, real inferences first
infreerence loot # everything that works, across all scans
infreerence serve # web UI + API → http://localhost:8000
```

---

## Commands

| Command | What it does |
|---|---|
| `scan` | search a provider's templates, then verify every hit |
| `verify` | probe host(s) you already have — no search, no credits |
| `keys` | find leaked provider API keys in a repo or folder, and validate them |
| `scans` / `show` | browse scan history and inspect one scan |
| `loot` | cross-scan inventory of everything found |
| `integrations` / `bridge` | turn a live endpoint into ready-to-paste tool config |
| `templates` / `providers` | browse the template corpus |
| `export` / `import` | move scans between machines as portable JSON |
| `serve` | run the web UI + API |

### Scan

```bash
infreerence scan ollama # one provider (all its templates)
infreerence scan ollama,litellm --limit all # several providers, one batch
infreerence scan all # every provider
infreerence scan ollama[ollama-running] -o out.json # a specific template
infreerence scan litellm --limit all --exhaustive # retrieve ALL of a huge result set
infreerence scan litellm --scope <CIDR> # restrict to in-scope targets
infreerence scan ollama --no-verify # search only, probe nothing
infreerence scan --resume ~/.infreerence/resume/<batch>.json # continue a Ctrl+C'd scan
```

`--limit all` pages until the source stops serving results; `--exhaustive` slices
a query so nothing is missed on very large sets. Ctrl+C writes a resume file and
prints the command to continue.
→ [Scanning in depth](wiki/Scanning.md) · [Stopping & resuming](wiki/Stop-and-Resume.md)

### Verify hosts you already have

```bash
infreerence verify <HOST>:4000 --provider litellm
infreerence verify <HOST>:4000 --provider all # which product IS this?
infreerence verify <HOST>:11434 <HOST-2>:8080 -p ollama
```

No search, so no query credits. → [Verifying hosts](wiki/Verify-Hosts.md)

### Find leaked API keys

```bash
infreerence keys owner/name                          # a GitHub repo
infreerence keys myorg/private-repo --account other-gh-user   # private, via gh auth
infreerence keys . ./src /path/to/checkout           # a working tree on disk
infreerence keys owner/name --reveal --json
```

Detects OpenAI, Anthropic and Moonshot/Kimi shapes, then asks each provider
whether the key still authenticates - the verdict, and the provider itself, come
from validation. → [API key scanning](wiki/API-Keys.md)

### Inspect what you found

```bash
infreerence scans # history
infreerence show <id> # detail + hits (real inferences)
infreerence show <id> --all --models # every hit, every model
infreerence loot # working findings, all scans
infreerence loot --kind apikey --all # every key ever found
infreerence loot -q claude --json # free-text search
```

Models are coloured by what actually answered: green works, red failed, amber slow
TPS, plain untested. → [Model verification](wiki/Model-Verification.md)

### Use what you found

```bash
infreerence integrations <scan_id> <HOST>:11434 # base URL + tool configs
infreerence integrations <scan_id> <HOST>:11434 --tool codex
infreerence bridge <scan_id> <HOST>:11434 --run # local LiteLLM bridge
```

→ [Integrations & the bridge](wiki/Integrations.md)

---

## Configuration

Credentials and tuning live in `.env`:

```ini
SHODAN_API_KEY=...                 # required for searching
VERIFY_MAX_NEW_CONNS_PER_SEC=8     # probe pacing — the setting that matters most
VERIFY_MAX_INFLIGHT=64             # concurrency ceiling
VERIFY_CHAT_TIMEOUT=90             # seconds per chat probe (slow local models)
VERIFY_HOST_DEADLINE=300           # seconds for all of one host's probing
VERIFY_MODEL_PROBE_MAX=8           # models probed per host
```

Probe pacing is the one to understand before a big scan: too high and your own
NAT/ISP drops connections, making live hosts look dead.
→ [Throughput & pacing](wiki/Throughput.md)

## Web UI & API

```bash
infreerence serve        # → http://localhost:8000  (OpenAPI docs at /docs)
```

Dashboard, template editor, scan launcher with live progress, loot inventory,
verify, integrations, and streaming chat with a discovered model.
→ [Web UI](wiki/Web-UI.md) · [HTTP API](wiki/API-Reference.md)

## Documentation

The full explanation of how the tool works lives in the [wiki](wiki/Home.md):

| Page | |
|---|---|
| [Concepts](wiki/Concepts.md) | templates, providers, sources, verifiers |
| [Scanning](wiki/Scanning.md) | `--limit all`, `--exhaustive`, batches, scope, honeypot memory |
| [Model verification](wiki/Model-Verification.md) | which models actually answer, and why some are amber |
| [API keys](wiki/API-Keys.md) | detection, validation, provider attribution, private repos |
| [Verifying hosts](wiki/Verify-Hosts.md) | `verify`, and identifying an unknown host |
| [Stopping & resuming](wiki/Stop-and-Resume.md) | what a resume actually re-does |
| [Throughput & pacing](wiki/Throughput.md) | why probing is rate-limited |
| [Integrations](wiki/Integrations.md) | Codex, Claude Code, Gemini CLI, the LiteLLM bridge |
| [Web UI](wiki/Web-UI.md) | every page |
| [HTTP API](wiki/API-Reference.md) | endpoints + the exchange format |
| [Development](wiki/Development.md) | source install, tests, standalone binary, layout |

## License

MIT — see [`LICENSE`](LICENSE).

infreerence

파일 탐색기

# CDN으로 사용하기

Was this content helpful?

Recent Feedback

Download README