Phase 0 Complete: Tasks 0.7-0.20
- 0.7: UI-Design-Richtlinien (docs/ui-design-guidelines.md, 535 lines) - 0.8: Theme-Customization Backend (4 theme fields, migration 0023) - 0.9: Theme-Customization Frontend (SettingsTheme.tsx, themeStore.ts, live preview) - 0.10: RBAC-Audit (4 plugins secured, 53 routes with require_permission) - 0.11: LiteLLM-Cleanup (llm_client.py migrated from httpx to litellm) - 0.12: KI-Agent-Framework docs (plugin-development-guide.md, agent_capabilities field) - 0.13: Heartbeat configurable (ProactiveSettings, migration 0024, frontend UI) - 0.14: Unified Search Field-Level RBAC (resolve_permissions + filter_fields_by_permission) - 0.15: Undo/History-System (EntityHistory model, service, routes, migration 0025, HistoryViewer) - 0.16: Storage Backend (LocalStorage + S3Storage, DMS/attachments/mail updated) - 0.17: Import/Export unified Contact fields (firstname, surname, email_1, phone_1) - 0.18: .gitignore & Config-Cleanup (webui→frontend, python-jose removed, .env untracked) - 0.19: Mail-Salt Security-Fix (per-account random salt, migration 0026) - 0.20: AGPL replaced (PyMuPDF→pypdf, OnlyOffice→Collabora, LICENSE + THIRD_PARTY_LICENSES.md)
This commit is contained in:
Agent Zero
+61
-36
@@ -1,8 +1,11 @@
|
||||
"""Configurable LLM client — supports OpenAI-compatible API or mock/stub mode.
|
||||
"""Configurable LLM client — supports LiteLLM (100+ providers) or mock/stub mode.
|
||||
|
||||
Reads AI_MODEL and AI_API_KEY from environment. If not set, uses mock mode
|
||||
Reads AI_MODEL, AI_API_KEY, AI_PROVIDER from environment. If not set, uses mock mode
|
||||
which returns predefined actions based on keyword matching. This allows
|
||||
tests to run without external API dependencies.
|
||||
|
||||
LiteLLM provides a unified interface to OpenAI, Anthropic, Google, Azure,
|
||||
AWS Bedrock, Ollama, and many more providers.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
@@ -12,7 +15,7 @@ import logging
|
||||
import os
|
||||
from typing import Any
|
||||
|
||||
import httpx
|
||||
import litellm
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
@@ -39,16 +42,20 @@ class LLMClient:
|
||||
"""LLM client that translates natural language to proposed API actions.
|
||||
|
||||
Modes:
|
||||
- If AI_MODEL and AI_API_KEY are set: calls OpenAI-compatible chat completions API
|
||||
- If AI_MODEL and AI_API_KEY are set: calls LiteLLM chat completions API
|
||||
- Otherwise: mock/stub mode with keyword-based action mapping
|
||||
|
||||
LiteLLM model format: "provider/model_name" (e.g. "openai/gpt-4o", "anthropic/claude-3-sonnet", "ollama/llama3")
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self, model: str | None = None, api_key: str | None = None, api_base: str | None = None
|
||||
):
|
||||
self, model: str | None = None, api_key: str | None = None, api_base: str | None = None,
|
||||
provider: str | None = None,
|
||||
) -> None:
|
||||
self.model = model or os.environ.get("AI_MODEL", "")
|
||||
self.api_key = api_key or os.environ.get("AI_API_KEY", "")
|
||||
self.api_base = api_base or os.environ.get("AI_API_BASE", "https://api.openai.com/v1")
|
||||
self.api_base = api_base or os.environ.get("AI_API_BASE", "")
|
||||
self.provider = provider or os.environ.get("AI_PROVIDER", "openai")
|
||||
self.is_mock = not bool(self.model and self.api_key)
|
||||
|
||||
async def generate(self, user_query: str, context: dict[str, Any] | None = None) -> LLMResponse:
|
||||
@@ -83,20 +90,28 @@ class LLMClient:
|
||||
)
|
||||
|
||||
async def _api_generate(self, query: str, context: dict[str, Any]) -> LLMResponse:
|
||||
"""Call OpenAI-compatible chat completions API.
|
||||
"""Call LLM via LiteLLM unified interface.
|
||||
|
||||
Sends a system prompt explaining the available API endpoints and asks
|
||||
the LLM to propose actions in structured JSON format.
|
||||
Supports 100+ providers through a single API:
|
||||
- OpenAI: "openai/gpt-4o"
|
||||
- Anthropic: "anthropic/claude-3-sonnet"
|
||||
- Google: "gemini/gemini-pro"
|
||||
- Azure: "azure/<deployment-name>"
|
||||
- Ollama: "ollama/llama3"
|
||||
- And many more.
|
||||
"""
|
||||
system_prompt = self._build_system_prompt(context)
|
||||
user_prompt = f"User request: {query}\n\nRespond with proposed actions as JSON."
|
||||
|
||||
headers = {
|
||||
"Authorization": f"Bearer {self.api_key}",
|
||||
"Content-Type": "application/json",
|
||||
}
|
||||
body = {
|
||||
"model": self.model,
|
||||
# Build LiteLLM model string: "provider/model" or just "model" for OpenAI compat
|
||||
if self.provider and self.provider != "openai":
|
||||
litellm_model = f"{self.provider}/{self.model}"
|
||||
else:
|
||||
litellm_model = self.model
|
||||
|
||||
# Build kwargs for litellm.acompletion
|
||||
kwargs: dict[str, Any] = {
|
||||
"model": litellm_model,
|
||||
"messages": [
|
||||
{"role": "system", "content": system_prompt},
|
||||
{"role": "user", "content": user_prompt},
|
||||
@@ -105,42 +120,52 @@ class LLMClient:
|
||||
"max_tokens": 1000,
|
||||
}
|
||||
|
||||
async with httpx.AsyncClient(timeout=30.0) as client:
|
||||
resp = await client.post(
|
||||
f"{self.api_base}/chat/completions",
|
||||
headers=headers,
|
||||
json=body,
|
||||
)
|
||||
resp.raise_for_status()
|
||||
data = resp.json()
|
||||
# Add API key if set
|
||||
if self.api_key:
|
||||
kwargs["api_key"] = self.api_key
|
||||
|
||||
content = data["choices"][0]["message"]["content"]
|
||||
return self._parse_llm_response(content)
|
||||
# Add API base if set (for self-hosted or custom endpoints)
|
||||
if self.api_base:
|
||||
kwargs["api_base"] = self.api_base
|
||||
|
||||
try:
|
||||
response = await litellm.acompletion(**kwargs)
|
||||
content = response.choices[0].message.content
|
||||
return self._parse_llm_response(content)
|
||||
except Exception as e:
|
||||
logger.error("LiteLLM API call failed: %s", e)
|
||||
# Fall back to mock mode on API error
|
||||
return LLMResponse(
|
||||
message=f"LLM API call failed: {e}. Falling back to keyword matching.",
|
||||
proposed_actions=[],
|
||||
confidence=0.1,
|
||||
)
|
||||
|
||||
def _build_system_prompt(self, context: dict[str, Any]) -> str:
|
||||
"""Build system prompt describing available API actions."""
|
||||
available_apis = [
|
||||
{"method": "GET", "path": "/api/v1/companies", "description": "List companies"},
|
||||
{"method": "POST", "path": "/api/v1/companies", "description": "Create a company"},
|
||||
{"method": "GET", "path": "/api/v1/contacts", "description": "List contacts (persons and companies)"},
|
||||
{"method": "POST", "path": "/api/v1/contacts", "description": "Create a contact (person or company)"},
|
||||
{
|
||||
"method": "GET",
|
||||
"path": "/api/v1/companies/{id}",
|
||||
"description": "Get company details",
|
||||
"path": "/api/v1/contacts/{id}",
|
||||
"description": "Get contact details",
|
||||
},
|
||||
{
|
||||
"method": "PATCH",
|
||||
"path": "/api/v1/companies/{id}",
|
||||
"description": "Update a company",
|
||||
"path": "/api/v1/contacts/{id}",
|
||||
"description": "Update a contact",
|
||||
},
|
||||
{
|
||||
"method": "DELETE",
|
||||
"path": "/api/v1/companies/{id}",
|
||||
"description": "Delete a company",
|
||||
"path": "/api/v1/contacts/{id}",
|
||||
"description": "Delete a contact",
|
||||
},
|
||||
{"method": "GET", "path": "/api/v1/contacts", "description": "List contacts"},
|
||||
{"method": "POST", "path": "/api/v1/contacts", "description": "Create a contact"},
|
||||
{"method": "GET", "path": "/api/v1/workflows", "description": "List workflows"},
|
||||
{"method": "POST", "path": "/api/v1/workflows", "description": "Create a workflow"},
|
||||
{"method": "GET", "path": "/api/v1/calendar/entries", "description": "List calendar entries"},
|
||||
{"method": "POST", "path": "/api/v1/calendar/entries", "description": "Create a calendar entry"},
|
||||
{"method": "GET", "path": "/api/v1/dms/files", "description": "List DMS files"},
|
||||
]
|
||||
context_str = json.dumps(context) if context else "{}"
|
||||
return (
|
||||
|
||||
Reference in New Issue
Block a user