Skip to content

A modern platform for visual, flexible, and extensible graph-based investigations. For cybersecurity analysts and investigators.

feat(core): add input and output type to template generator

dextmorgn6b35cccddbee
diff --git a/flowsint-core/src/flowsint_core/core/services/template_generator_service.py b/flowsint-core/src/flowsint_core/core/services/template_generator_service.py
index 66945d4d..0ed03e56 100644
--- a/flowsint-core/src/flowsint_core/core/services/template_generator_service.py
+++ b/flowsint-core/src/flowsint_core/core/services/template_generator_service.py
@@ -4,8 +4,10 @@ Service for AI-assisted enricher template generation.
Uses an LLM to generate valid enricher template YAML from a free-text prompt.
"""
+import json
import os
import re
+from typing import Any, Dict, Optional
from uuid import UUID
import yaml
@@ -163,12 +165,49 @@ class TemplateGeneratorService(BaseService):
api_key = self._vault_service.get_secret(owner_id, vault_key)
return create_llm_provider(provider=provider_name, api_key=api_key)
- async def generate(self, prompt: str, owner_id: UUID) -> str:
+ def _build_type_context(
+ self,
+ input_type: Optional[str],
+ input_schema: Optional[Dict[str, Any]],
+ output_type: Optional[str],
+ output_schema: Optional[Dict[str, Any]],
+ ) -> str:
+ """Build additional context about input/output type schemas."""
+ parts: list[str] = []
+ if input_type and input_schema:
+ parts.append(
+ f"## Input type: {input_type}\n"
+ f"The template MUST use `input.type: {input_type}`.\n"
+ f"Schema (available fields on the input):\n"
+ f"```json\n{json.dumps(input_schema, indent=2)}\n```"
+ )
+ if output_type and output_schema:
+ parts.append(
+ f"## Output type: {output_type}\n"
+ f"The template MUST use `output.type: {output_type}`.\n"
+ f"The `response.map` keys MUST only use fields from this schema:\n"
+ f"```json\n{json.dumps(output_schema, indent=2)}\n```"
+ )
+ return "\n\n".join(parts)
+
+ async def generate(
+ self,
+ prompt: str,
+ owner_id: UUID,
+ input_type: Optional[str] = None,
+ input_schema: Optional[Dict[str, Any]] = None,
+ output_type: Optional[str] = None,
+ output_schema: Optional[Dict[str, Any]] = None,
+ ) -> str:
"""Generate an enricher template YAML from a free-text description.
Args:
prompt: User's free-text description of the desired enricher.
owner_id: ID of the user (for vault-based LLM API key).
+ input_type: Name of the input Flowsint type (e.g. "Ip").
+ input_schema: JSON schema of the input type.
+ output_type: Name of the output Flowsint type (e.g. "SocialAccount").
+ output_schema: JSON schema of the output type.
Returns:
Raw YAML string of the generated template.
@@ -179,8 +218,20 @@ class TemplateGeneratorService(BaseService):
"""
provider = self._get_llm_provider(owner_id)
+ system_content = _SYSTEM_PROMPT
+ type_context = self._build_type_context(
+ input_type, input_schema, output_type, output_schema
+ )
+ if type_context:
+ system_content += (
+ "\n\n## Type Constraints (from the user's selection)\n\n"
+ + type_context
+ + "\n\nYou MUST respect the input and output types above. "
+ "Only use fields that exist in the provided schemas for the response.map keys."
+ )
+
messages = [
- ChatMessage(role=MessageRole.SYSTEM, content=_SYSTEM_PROMPT),
+ ChatMessage(role=MessageRole.SYSTEM, content=system_content),
ChatMessage(role=MessageRole.USER, content=prompt),
]