DeepTeam is a framework to red team LLMs and AI agents.
updated docs
penguine582e18f521d2
diff --git a/docs/docs/guardrails-introduction.mdx b/docs/docs/guardrails-introduction.mdxindex e3551148..cf5e048e 100644--- a/docs/docs/guardrails-introduction.mdx+++ b/docs/docs/guardrails-introduction.mdx@@ -63,13 +63,12 @@ guardrails = Guardrails()```-There are **TWO** mandatory and **TWO** optional parameters when creating a `Guardrails`:- `input_guards`: a list of input guards for protecting against malicious inputs- `output_guards`: a list of output guards for evaluating LLM outputs-- [Optional] `evaluation_model`: a string specifying which OpenAI model to use for guard evaluation (defaults to `"gpt-4.1"`)-- [Optional] `sample_rate`: a float between 0.0 and 1.0 that determines what fraction of requests are guarded (defaults to `1.0`)+- [Optional] `evaluation_model`: a string specifying which OpenAI model to use for guard evaluation. Defaulted to `gpt-4.1`.+- [Optional] `sample_rate`: a float between 0.0 and 1.0 that determines what fraction of requests are guarded. Defaulted to `1.0`.```python# Advanced configuration@@ -80,6 +79,7 @@ guardrails = Guardrails(sample_rate=0.5 # Guard 50% of requests)```+:::noteYour list of `input_guards` and `output_guards` must not be empty if you wish to guard the respecitve inputs and outputs.:::@@ -98,26 +98,6 @@ print(guard_result.breached)The `guard_input` method will invoke every single guard within your `Guardrails`s to check whether a particular input has breached any of the defined criteria.-### 3-Tier Safety Assessment--Each guard uses a **3-tier safety assessment system**:--- **`safe`**: The content is clearly safe and poses no risk-- **`uncertain`**: The content is borderline or ambiguous, requiring human review-- **`unsafe`**: The content clearly violates safety guidelines--```python-# Access detailed safety assessment-print(guard_result.safety_level) # "safe", "uncertain", or "unsafe"-print(guard_result.reason) # Explanation for the assessment-```--A `Guardrails` is considered **breached** if any guard returns `unsafe` or `uncertain`.--:::tip-`Guardrails` used against malicious inputs is the best way to not waste tokens generating text that you shouldn't even be generating in the first place.-:::-### Guard an OutputSimiarly, call the `guard_output` method wherever appropriate in your code to guard against an LLM output:@@ -167,4 +147,22 @@ Output guards evaluate your LLM's responses before they reach users, ensuring sa- **[Hallucination Guard](/docs/guardrails-hallucination-guard)** - Detects and prevents fabricated, inaccurate, or hallucinated information in responses- **[Cybersecurity Guard](/docs/guardrails-cybersecurity-guard)** - (When configured for output) Ensures responses don't contain dangerous cybersecurity information or instructions+## 3-Tier Safety Assessment++All of `deepteam`'s guards uses a **3-tier safety assessment system**:++- `safe`: The content is clearly safe and poses no risk+- `uncertain`: The content is borderline or ambiguous, requiring human review+- `unsafe`: The content clearly violates safety guidelines+```python+# Access detailed safety assessment+print(guard_result.safety_level) # "safe", "uncertain", or "unsafe"+print(guard_result.reason) # Explanation for the assessment+```++A `Guardrails` is considered **breached** if any guard returns `unsafe` or `uncertain`.++:::tip+`Guardrails` used against malicious inputs is the best way to not waste tokens generating text that you shouldn't even be generating in the first place.+:::diff --git a/docs/sidebars.js b/docs/sidebars.jsindex 150a973c..dc207478 100644--- a/docs/sidebars.js+++ b/docs/sidebars.js@@ -142,51 +142,26 @@ module.exports = {{type: "category",label: "Guardrails",- items: ["guardrails-introduction", {- type: "category",- label: "Guards",- items: [ "guardrails-prompt-injection-guard", "guardrails-toxicity-guard"],- collapsed: false,- }],+ className: "sidebar-item-icon-guardrails",+ items: [+ "guardrails-introduction",+ {+ type: "category",+ label: "Guards",+ items: [+ "guardrails-prompt-injection-guard",+ "guardrails-topical-guard",+ "guardrails-toxicity-guard",+ "guardrails-privacy-guard",+ "guardrails-illegal-guard",+ "guardrails-hallucination-guard",+ "guardrails-cybersecurity-guard",+ ],+ collapsed: true,+ },+ ],collapsed: false,},- // {- // type: "category",- // label: "Guardrails",- // className: "sidebar-item-icon-guardrails",- // items: [- // "guardrails-introduction",- // {- // type: "category",- // label: "Input Guards",- // items: [- // "guardrails-prompt-injection-guard",- // "guardrails-topical-guard",- // ],- // collapsed: true,- // },- // {- // type: "category",- // label: "Output Guards",- // items: [- // "guardrails-toxicity-guard",- // "guardrails-privacy-guard",- // "guardrails-illegal-guard",- // "guardrails-hallucination-guard",- // ],- // collapsed: true,- // },- // {- // type: "category",- // label: "Input/Output Guards",- // items: [- // "guardrails-cybersecurity-guard",- // ],- // collapsed: true,- // },- // ],- // collapsed: false,- // },{type: "category",label: "Guidelines & Frameworks",