LLM-Shield-Proxy is a self-hosted gateway for OpenAI-compatible LLM APIs. It replaces the personal
data and secrets it detects (emails, card numbers, SSNs, API keys and more) before a request goes
to the model provider, and puts the original values back into the streamed response before your
application sees it. Your application changes only its base_url and the key it sends.
Already running LiteLLM? LiteLLM includes it as a built-in guardrail, guardrail: llm_shield_proxy,
so LiteLLM keeps its own routing and calls the Shield to redact each request and restore each reply.
See LiteLLM's setup page. It is
on LiteLLM's main branch and not yet in a tagged LiteLLM release; until then, use one of the
wirings in Running behind LiteLLM.
On a LiteLLM without the module, that config does not stop the proxy: LiteLLM logs one
error line, Skipping guardrail 'llm-shield': invalid configuration, proxy is starting WITHOUT this guardrail: Unsupported guardrail: llm_shield_proxy, starts, and every request reaches the
provider unredacted (seen on 1.105.0, the newest image). Check that line is absent from the
startup log before sending anything real.
pip install llm-shield-proxy
UPSTREAM_BASE_URL=http://127.0.0.1:8765 UPSTREAM_API_KEY=unused VALID_VIRTUAL_KEYS=sk-demo llm-shield-proxy --port 4000 &
pii-leak-benchmark selfcheck --target-base-url http://127.0.0.1:4000/v1 --target-api-key sk-demoThe second command starts the proxy. The third sends prompts full of synthetic emails, SSNs, card
numbers and API keys through it, and plays the model provider at 127.0.0.1:8765, so it can see
exactly what the proxy forwarded. Nothing calls a real model. You should see:
CLEAN
No fixture value reached the upstream; required restore checks passed.
Data types tested
TYPE RESULT WHAT IT MEANS
AWS_ACCESS_KEY_ID contained never reached the upstream in this run
CREDIT_CARD contained never reached the upstream in this run
EMAIL contained never reached the upstream in this run
GITHUB_TOKEN contained never reached the upstream in this run
SLACK_TOKEN contained never reached the upstream in this run
SSN contained never reached the upstream in this run
Run the same check with nothing in the middle (--target-base-url capture://self) and every row
reads LEAK. On Windows PowerShell, use the
PowerShell version of these commands.
The proxy needs two keys. VALID_VIRTUAL_KEYS lists the keys your clients send to the proxy;
any other key gets a 401. The provider key (OPENAI_API_KEY here) is what the proxy sends
upstream, so your clients never hold it.
pip install llm-shield-proxy
export VALID_VIRTUAL_KEYS=sk-my-client-key
export OPENAI_API_KEY=sk-your-openai-key
llm-shield-proxy --host 127.0.0.1 --port 8000Then point your existing client at it:
from openai import OpenAI
client = OpenAI(api_key="sk-my-client-key", base_url="http://localhost:8000/v1")
stream = client.chat.completions.create(
model="gpt-4o-mini",
messages=[{"role": "user", "content": "Contact Sarah at sarah@example.com."}],
stream=True,
)
for chunk in stream:
print(chunk.choices[0].delta.content or "", end="")For another provider, set UPSTREAM_BASE_URL and that provider's key; every setting is described
in .env.example and the deployment guide.
With Docker, docker-compose.yml sets both keys for you. Put your OpenAI key in OPENAI_API_KEY
first; the demo sends the client key demo-key:
docker compose up -d
curl http://localhost:8000/healthz
python examples/demo.py
Before sending a request to the model provider, the proxy finds configured types of sensitive data and replaces their values. As the provider streams its response, the proxy joins replacement tokens that were split across SSE events and restores values that the client is allowed to receive. For structured JSON, it changes string values without changing the JSON syntax. Test this behavior with the schemas used by your provider and tools.
The maintained component map and deployment diagrams live in the architecture guide, architecture whitepaper, and deployment guide.
| Area | What is implemented | Where the evidence stops |
|---|---|---|
| Detection | 11 native Tier 1 data types, Tier 2 Shannon entropy, optional Tier 3 ONNX NER, BYOR rules | Supported types · no recall guarantee on unlabeled traffic |
| Streaming privacy | Sliding-window SSE rehydration, bounded streaming JSON lexer | Architecture · conformance method |
| Masking | Synthetic, structural-tag, scrub, operator-keyed stateless crypto | Masking guide · plaintext still exists in process memory |
| Security controls | SSRF/DNS-rebinding egress checks, request policy, rate and blast-radius limits, canary tripwires | Security · not a substitute for network policy |
| Evidence plane | Hash-linked audit records, Ed25519 receipts, OSCAL output, compliance packs | Compliance overview · tamper-evident, not WORM without immutable retention |
| MCP governance | Scoped JSON-RPC subset with RBAC and egress policy | Research-scoped; MCP guide · not a complete MCP transport |
In standard mode, detection, masking, policy checks, and value restoration run inside your gateway. Only the masked request is sent to the external model provider:
In air-gapped mode, the masked request goes to an internal model gateway. Network policy must still block direct provider access, telemetry, and other unintended outbound traffic:
See deployment topologies, air-gapped egress, and the Kubernetes/Helm deployment guide.
Every feature is labelled Supported, Beta, Experimental, or Research. The label states how
the feature was tested and what remains untested: feature catalog ·
stability policy · limitations.
It supports SOC 2, HIPAA, GDPR, EU AI Act and NIST/ISO evidence programs by supplying technical controls and artifacts. It does not certify a deployment, guarantee complete detection, or make network policy optional.
The proxy is one of four packages here. The other three are standalone and do not install it:
pii-leak-benchmarktests an OpenAI-compatible streaming gateway. It checks whether the gateway sends the test values to its model provider and whether the client gets the original values back.mcp-ssrf-checkchecks an MCP server you operate for missingHostandOriginvalidation, unbound session ids, and URL-fetching tools that reach loopback. It talks only to your server and to a listener it opens on your own machine. It also runs as a GitHub Action that writes the result table to the job summary.chunk-invarianceturns the split-boundary rule into one test assertion: every way of splitting an input into chunks must stream to the same output as filtering it whole. Python on PyPI, and TypeScript on npm (chunk-invariance-js/).
pip install pii-leak-benchmark
# The negative control: no gateway at all, raw pass-through. MUST report outcome=fail.
pii-leak-benchmark \
--target-base-url capture://self \
--target-name raw-pass-through-negative-control --target-version 1 \
--redaction-claimed claimed \
--redaction-claim-citation https://github.com/ninadphalak/LLM-Shield-Proxy/blob/main/website/docs/conformance/reproducing.md \
--redaction-enabled \
--redaction-config-reference "synthetic control: declared redaction intentionally absent"
# Your gateway, already configured to send upstream traffic to http://127.0.0.1:8765/v1
pii-leak-benchmark --target-base-url http://127.0.0.1:4000/v1 --target-name your-gatewayOne bounded experiment, offline, no gateway or account. It re-runs the chunk-local and length-bounded-retention inspectors at the published seed and diffs every field of the result against the published reports:
git clone https://github.com/ninadphalak/LLM-Shield-Proxy.git
cd LLM-Shield-Proxy
python -m pip install ./pii-leak-benchmark
python benchmarks/reproduce_fragmentation.py --out reproductionAbout two minutes. Exit status 0 means every field matched except timestamps and wall-clock
timings. The same command runs in CI on Ubuntu, macOS and Windows across Python 3.11 and 3.12
(fragmentation-reproduction in .github/workflows/benchmark.yml).
Full walkthrough and what the numbers mean:
reproduce the fragmentation result.
One question, one command: does your deployment send raw personal data to its upstream?
pip install "pii-leak-benchmark>=0.2.0"
# Establish the floor first. No gateway at all: this MUST report LEAK.
pii-leak-benchmark selfcheck --target-base-url capture://self
# Then your own gateway, already configured to use the capture as its upstream.
pii-leak-benchmark selfcheck --target-base-url http://your-gateway.internal/v1selfcheck needs 0.2.0 or newer.
To fail your own build on a leak, the complete workflow file and the three settings you have
to change are in the CI setup guide, with a
shorter copy in
examples/ci/gateway-pii-check.yml. That path needs
0.3.0 or newer, which adds the ci subcommand, the GitHub Action, and baseline comparison
against your previous version.
To build the proxy in its own fork and retain both request and 32-case response reports, use the source-build CI recipes for LLM-Shield-Proxy, Portkey, LiteLLM, or NeMo Guardrails. The workflow starts its own synthetic capture, so no model account is needed.
It reports one row per data type, so you can see what your gateway handled and what it missed:
TYPE RESULT WHAT IT MEANS
AWS_ACCESS_KEY_ID LEAK sent to the upstream unmasked
CREDIT_CARD contained never reached the upstream in this run
EMAIL LEAK sent to the upstream unmasked
GITHUB_TOKEN LEAK sent to the upstream unmasked
SLACK_TOKEN LEAK sent to the upstream unmasked
SSN contained never reached the upstream in this run
That is a real result from a gateway that redacts SSNs and card numbers and forwards every credential, which is the common shape. Six types: email, SSN, card, plus AWS, GitHub and Slack credential specimens. Every run also prints what it did not test, because the dangerous reading of a clean result is "my gateway handles sensitive data".
Exit 0 CLEAN, 1 LEAK, 2 NOT MEASURED. The third is the one that matters: a gateway that
answers your client but was never pointed at the capture inspects nothing, so every check passes
vacuously. That gets its own exit code instead of reading as a pass. selfcheck requires no vendor
claim and its report is deliberately not publishable as a row about a product; to publish a
comparative result, use the flat command and record the claim.
The only third-party Python dependency is httpx; you do not need to install one gateway to test
another. You configure the gateway to use the benchmark's local capture server as its model
provider. The benchmark then checks the URL, headers, HTTP framing, and JSON body for the test values. If
it cannot safely parse part of the request, the run ends with an error instead of assuming that no
value leaked. The conformance docs describe the full method.
fail has one narrow meaning: the gateway sent an unmasked test value to the benchmark's capture
server. A product that does not offer PII redaction is marked not-applicable, not failed. A
one-way anonymizer that removes the values but does not restore them receives a separate outcome.
| Target | Outcome | Runs / distinct submitters |
|---|---|---|
| Raw capture endpoint (control) | fail - three literal matches |
1 / 1 - control, not a product |
| LLM-Shield-Proxy | pass - 5/5 |
1 / 1 - unreplicated |
| LiteLLM 1.99.0, default | redaction-not-enabled |
1 / 1 - unreplicated |
| LiteLLM 1.99.0 + Presidio | no-leak-profile-not-met (no leak) |
1 / 1 - unreplicated |
| Portkey OSS 1.15.2, default | redaction-not-enabled |
1 / 1 - unreplicated |
| Portkey OSS 1.15.2 + regexReplace | no-leak-profile-not-met (no leak) |
1 / 1 - unreplicated |
Every product result above was run once by this project's maintainer. It has not yet been repeated
by an independent person. A result becomes replicated only after three different people each
submit a run of the same gateway and configuration. Until then, it remains unreplicated.
Full table, method and evidence ·
submit a run.
The six results below were produced by this project on one workstation. No outside contributor has
repeated them yet, so the table marks every product result as unreplicated. Each result links to
the exact configuration and the report produced by the run.
The first version of the test used invalid examples of an email address, SSN, and credit card:
| Old test value | Why Presidio rejected it |
|---|---|
person@example.invalid |
.invalid is not a public domain suffix |
123-45-6789 |
Presidio blocks this well-known invalid SSN sequence |
4532-1234-5678-9012 |
The number fails the Luhn card-number checksum |
LLM-Shield-Proxy matched the text patterns but did not perform those validity checks. This gave it
an unfair advantage over detectors that validate values. A LiteLLM and Presidio run revealed the
problem: the old values produced leaked: ["SSN"], while valid test values produced leaked: [].
The project did not publish the affected result. It replaced the three values with valid, reserved
test values and reran all six configurations. See the
fixture threat model for the full record.
The benchmark also found two streaming bugs in LLM-Shield-Proxy. It created an OpenTelemetry span
for every SSE event even when export was off, and it sent each event's blank terminator as a
separate write. Both bugs are fixed and covered by
tests/test_streaming_write_efficiency.py. The
run record explains what changed.
Known limitation: the test uses three fixed data formats. A small program written specifically for those formats can pass without being a general PII detector. The values change on every run, but the formats do not. Testing more formats caused two false failures in six trials. See the fixture threat model for the measurements.
git clone https://github.com/ninadphalak/LLM-Shield-Proxy.git
cd LLM-Shield-Proxy
python -m pip install -e ./pii-leak-benchmark -e ".[dev]"
python -m pytestThe benchmark is a separate distribution in this repo, so it installs first; nothing in it imports the proxy and a test fails if that ever changes. CI provisions real Redis, an HTTP/2 ALPN server, a checksum-pinned ONNX export, Docker, Helm and promtool. A missing dependency fails those jobs rather than skipping them, so a green build cannot mean "nothing ran".
- Start here: interactive docs and playground · configuration · deployment · operations · troubleshooting
- Understand the design: architecture · architecture whitepaper · feature catalog · stability · limitations
- Integrate it: integration index · running behind LiteLLM · LiteLLM and Ollama recipe · Open WebUI and LangChain recipe · migration from Presidio
- Operate securely: security model · policy as code · compliance evidence mapping · immutable retention
- Verify and participate: conformance method and results · submit a reproduction · 30-day design-partner pilot
Contributions are welcome through issues, discussions and CONTRIBUTING.md. The most valuable contribution is an independent benchmark run against a gateway you operate, whether it matches or differs from a row above.
Source code is Apache 2.0, which includes a patent license for using this code; documentation and diagrams may carry CC BY 4.0 terms. See LICENSE and NOTICE.
If you reference the architecture or benchmark methodology, use CITATION.cff or:
Phalak, N. (2026). Quantifying Latency and Token Overhead in Real-Time LLM Stream Sanitization: A Tiered Detection Approach. https://doi.org/10.5281/zenodo.21955770