llm-firewall

Name: llm-firewall
Author: CollieAi

43stars1forks

How to install llm-firewall?

collieai/llm-firewall

$git clone https://github.com/CollieAi/llm-firewall

Installs into the current project.

Summary

You're building with GPT-4, Claude, Gemini — but every API call is a potential data leak. Users paste credit cards into chat. Attackers inject malicious prompts. Sensitive data flows to third-party models without any filtering.

Files · 1

View on GitHub

README.md

1	<p align="center">
2	<a href="https://collieai.io">
3	<img src="https://collieai.io/logo-1.svg" alt="CollieAi" width="80" />
4	</a>
5	</p>
6
7	<h1 align="center">LLM Firewall — AI Security Proxy & Guardrails for GPT, Claude, Gemini</h1>
8
9	<h4 align="center">
10	Drop-in, provider-agnostic AI guardrails proxy that masks PII, blocks prompt injection & jailbreaks, and logs every LLM call.
11	One line change. Works with OpenAI, Anthropic Claude, Google Gemini, Azure OpenAI, AWS Bedrock, and self-hosted models.
12	</h4>
13
14	<p align="center">
15	<a href="https://collieai.io"><img src="https://img.shields.io/badge/website-collieai.io-blue" alt="CollieAi Website" /></a>
16	<a href="https://docs.collieai.io"><img src="https://img.shields.io/badge/docs-docs.collieai.io-green" alt="Documentation" /></a>
17	<a href="https://app.collieai.io"><img src="https://img.shields.io/badge/free_tier-20K_calls/month-brightgreen" alt="Free Tier" /></a>
18	</p>
19
20	<p align="center">
21	<a href="https://docs.collieai.io/getting-started/quick-start">Quick Start</a> · <a href="https://docs.collieai.io">Docs</a> · <a href="https://docs.collieai.io/api-reference">API Reference</a> · <a href="https://app.collieai.io">Try Free</a> · <a href="https://docs.collieai.io/self-hosting">Self-Host</a>
22	</p>
23
24	## Why LLM Firewall?
25
26	You're building with GPT-4, Claude, Gemini — but every API call is a potential data leak. Users paste credit cards into chat. Attackers inject malicious prompts. Sensitive data flows to third-party models without any filtering.
27
28	LLM Firewall by [CollieAi](https://collieai.io) is a drop-in security proxy that sits between your application and the LLM provider. It inspects, masks, and blocks — in real time, with zero code changes beyond swapping the base URL.
29
30	<!-- GIF: dashboard, policy engine, real-time logs -->
31
32	<img width="1000" height="561" alt="CollieAi LLM firewall blocking a prompt injection attack in real time — demo" src="https://github.com/user-attachments/assets/de356d6e-6483-4d30-aa80-d8c36482816e" />
33
34
35
36	## Key Features
37
38	- Drop-in OpenAI-compatible proxy — change `base_url`, keep your SDK. Python, Node.js, Go, curl — anything that speaks OpenAI works with CollieAi
39	- PII detection & masking — regex patterns, structured ID validation (Luhn, MOD-97), and dictionary matching via Aho-Corasick
40	- Prompt injection protection — ML classifier (10-50 ms) plus optional LLM-based analysis for sophisticated attacks
41	- Flexible policy engine — 6 configurable rule categories, each with monitor (log only) or enforce (mask/block) mode. Custom regex, dictionaries, thresholds — tune everything per project
42	- Input + output filtering — scan both user prompts and model responses with independent policies
43	- User satisfaction tracking — detect negative sentiment, monitor dissatisfaction metrics, get alerts when something goes wrong
44	- Full audit trail — every request logged with triggered rules, latency, and token counts. Filter by project, time range, or rule type
45	- Dashboard — create projects, configure rules, add provider tokens, review logs. No config files, no deployments
46
47	> Works with: OpenAI · Claude · deepseek · Gemini · Azure OpenAI · AWS Bedrock · any OpenAI-compatible endpoint
48
49	## Quick Start — 3 Steps
50
51	<table>
52	<tr>
53	<td width="40%">
54
55	1. Point your client at CollieAi
56
57	Sign up for a free API key and change your `base_url`. No SDK required, no architecture changes.
58
59	</td>
60	<td width="60%">
61
62	```python
63	# Change two config values
64	base_url = "https://app.collieai.io/v1"
65	api_key = "clai_your_project_key"
66
67	# Free: 20,000 calls/month
68	```
69
70	</td>
71	</tr>
72	<tr>
73	<td width="40%">
74
75	2. Your code stays the same
76
77	CollieAi is a transparent proxy. Use the standard OpenAI SDK — every request is filtered by your policy rules automatically.
78
79	</td>
80	<td width="60%">
81
82	```python
83	from openai import OpenAI
84
85	client = OpenAI(
86	base_url="https://app.collieai.io/v1",
87	api_key="clai_...",
88	)
89	client.chat.completions.create(...)
90	```
91
92	</td>
93	</tr>
94	<tr>
95	<td

People who liked this also viewed

6 picks

Type

llm-firewall

43stars1forks

View on GitHub

How to install llm-firewall?

collieai/llm-firewall

$git clone https://github.com/CollieAi/llm-firewall

Installs into the current project.

Summary

Files · 1

View on GitHub

README.md

1	<p align="center">
2	<a href="https://collieai.io">
3	<img src="https://collieai.io/logo-1.svg" alt="CollieAi" width="80" />
4	</a>
5	</p>
6
7	<h1 align="center">LLM Firewall — AI Security Proxy & Guardrails for GPT, Claude, Gemini</h1>
8
9	<h4 align="center">
10	Drop-in, provider-agnostic AI guardrails proxy that masks PII, blocks prompt injection & jailbreaks, and logs every LLM call.
11	One line change. Works with OpenAI, Anthropic Claude, Google Gemini, Azure OpenAI, AWS Bedrock, and self-hosted models.
12	</h4>
13
14	<p align="center">
15	<a href="https://collieai.io"><img src="https://img.shields.io/badge/website-collieai.io-blue" alt="CollieAi Website" /></a>
16	<a href="https://docs.collieai.io"><img src="https://img.shields.io/badge/docs-docs.collieai.io-green" alt="Documentation" /></a>
17	<a href="https://app.collieai.io"><img src="https://img.shields.io/badge/free_tier-20K_calls/month-brightgreen" alt="Free Tier" /></a>
18	</p>
19
20	<p align="center">
21	<a href="https://docs.collieai.io/getting-started/quick-start">Quick Start</a> · <a href="https://docs.collieai.io">Docs</a> · <a href="https://docs.collieai.io/api-reference">API Reference</a> · <a href="https://app.collieai.io">Try Free</a> · <a href="https://docs.collieai.io/self-hosting">Self-Host</a>
22	</p>
23
24	## Why LLM Firewall?
25
26	You're building with GPT-4, Claude, Gemini — but every API call is a potential data leak. Users paste credit cards into chat. Attackers inject malicious prompts. Sensitive data flows to third-party models without any filtering.
27
28	LLM Firewall by [CollieAi](https://collieai.io) is a drop-in security proxy that sits between your application and the LLM provider. It inspects, masks, and blocks — in real time, with zero code changes beyond swapping the base URL.
29
30	<!-- GIF: dashboard, policy engine, real-time logs -->
31
32	<img width="1000" height="561" alt="CollieAi LLM firewall blocking a prompt injection attack in real time — demo" src="https://github.com/user-attachments/assets/de356d6e-6483-4d30-aa80-d8c36482816e" />
33
34
35
36	## Key Features
37
38	- Drop-in OpenAI-compatible proxy — change `base_url`, keep your SDK. Python, Node.js, Go, curl — anything that speaks OpenAI works with CollieAi
39	- PII detection & masking — regex patterns, structured ID validation (Luhn, MOD-97), and dictionary matching via Aho-Corasick
40	- Prompt injection protection — ML classifier (10-50 ms) plus optional LLM-based analysis for sophisticated attacks
41	- Flexible policy engine — 6 configurable rule categories, each with monitor (log only) or enforce (mask/block) mode. Custom regex, dictionaries, thresholds — tune everything per project
42	- Input + output filtering — scan both user prompts and model responses with independent policies
43	- User satisfaction tracking — detect negative sentiment, monitor dissatisfaction metrics, get alerts when something goes wrong
44	- Full audit trail — every request logged with triggered rules, latency, and token counts. Filter by project, time range, or rule type
45	- Dashboard — create projects, configure rules, add provider tokens, review logs. No config files, no deployments
46
47	> Works with: OpenAI · Claude · deepseek · Gemini · Azure OpenAI · AWS Bedrock · any OpenAI-compatible endpoint
48
49	## Quick Start — 3 Steps
50
51	<table>
52	<tr>
53	<td width="40%">
54
55	1. Point your client at CollieAi
56
57	Sign up for a free API key and change your `base_url`. No SDK required, no architecture changes.
58
59	</td>
60	<td width="60%">
61
62	```python
63	# Change two config values
64	base_url = "https://app.collieai.io/v1"
65	api_key = "clai_your_project_key"
66
67	# Free: 20,000 calls/month
68	```
69
70	</td>
71	</tr>
72	<tr>
73	<td width="40%">
74
75	2. Your code stays the same
76
77	CollieAi is a transparent proxy. Use the standard OpenAI SDK — every request is filtered by your policy rules automatically.
78
79	</td>
80	<td width="60%">
81
82	```python
83	from openai import OpenAI
84
85	client = OpenAI(
86	base_url="https://app.collieai.io/v1",
87	api_key="clai_...",
88	)
89	client.chat.completions.create(...)
90	```
91
92	</td>
93	</tr>
94	<tr>
95	<td

People who liked this also viewed

6 picks

Type