agno-deepkeep

DeepKeep AI Firewall custom guardrails for Agno.

This package follows Agno's native guardrail extension point. Add DeepKeepGuardrail to an Agent or Team with pre_hooks to check input before it reaches the model, or with post_hooks to check output before it is returned.

Installation

pip install agno-deepkeep

Set credentials:

export DEEPKEEP_API_KEY="dk_..."
export DEEPKEEP_BASE_URL="https://api.example.deepkeep.ai"

Input guardrails

Use pre_hooks to run DeepKeep pre-moderation before an Agno Agent or Team processes user input:

from agno.agent import Agent
from agno_deepkeep import DeepKeepGuardrail

agent = Agent(
    name="DeepKeep Protected Agent",
    model="openai:gpt-5.2",
    pre_hooks=[
        DeepKeepGuardrail(
            pre_model="input-firewall-id",
        )
    ],
)

Output guardrails

Use post_hooks to run DeepKeep post-moderation after the model response is generated:

from agno.agent import Agent
from agno_deepkeep import DeepKeepGuardrail

agent = Agent(
    name="DeepKeep Protected Agent",
    model="openai:gpt-5.2",
    post_hooks=[
        DeepKeepGuardrail(
            post_model="output-firewall-id",
        )
    ],
)

Input and output protection

Use separate instances when you want distinct firewall IDs for input and output checks:

from agno.agent import Agent
from agno_deepkeep import DeepKeepGuardrail

agent = Agent(
    name="DeepKeep Protected Agent",
    model="openai:gpt-5.2",
    pre_hooks=[DeepKeepGuardrail(pre_model="input-firewall-id")],
    post_hooks=[DeepKeepGuardrail(post_model="output-firewall-id")],
)

DeepKeep endpoints

This package uses the same DeepKeep OpenAI-compatible moderation endpoints as the LangChain integration:

• POST /api/v3/openai/moderations/pre with {"model", "input", "title", "chat"}.
• POST /api/v3/openai/moderations/post with {"model", "output", "title", "chat"}.
• X-API-Key for authentication.

pre_model and post_model are DeepKeep firewall IDs. Each value is sent to the matching moderation endpoint as the model field.

Guardrail actions

DeepKeep responses are interpreted from verbosity[].details.guardrail_action:

• block: Raises Agno InputCheckError or OutputCheckError.
• redact or modify: Replaces Agno input or output when DeepKeep returns modified content.
• alert: Allows the run to continue.

If DeepKeep returns an unsupported action, the guardrail fails closed.

Configuration

• api_key: DeepKeep API key. If omitted, reads DEEPKEEP_API_KEY.
• base_url: DeepKeep base URL, without a trailing slash. If omitted, reads DEEPKEEP_BASE_URL.
• pre_model: DeepKeep firewall ID for pre-moderation.
• post_model: DeepKeep firewall ID for post-moderation.
• title: Optional title value sent to DeepKeep.
• chat: Optional chat value sent to DeepKeep.
• timeout: Request timeout in seconds. Defaults to 30.0.
• fail_closed: Raise a guardrail error when DeepKeep cannot be reached. Defaults to True.

Recommended Agno docs positioning

Position the public integration as DeepKeep AI Firewall in Agno's custom guardrails area:

DeepKeep AI Firewall
Use DeepKeep AI Firewall as custom guardrails for Agno Agents and Teams. Add DeepKeepGuardrail to pre_hooks or post_hooks to enforce runtime policies for prompt injection, jailbreaks, sensitive data leakage, unsafe content, and other AI risks.

For production enforcement, prefer Agno guardrails over agent-callable tools because guardrails run deterministically before or after model execution.