Mise en place docker + mise en place des settings (config ollama / 1min.ai)
This commit is contained in:
@@ -5,8 +5,9 @@ au domaine via injection de dépendance (ports + use cases), et transforme les
|
||||
erreurs du domaine en réponses HTTP. Aucune connaissance d'Ollama ici.
|
||||
"""
|
||||
import json
|
||||
from typing import Annotated, AsyncIterator
|
||||
from typing import Annotated, AsyncIterator, Literal
|
||||
|
||||
import httpx
|
||||
from fastapi import Depends, FastAPI, HTTPException
|
||||
from fastapi.responses import StreamingResponse
|
||||
from pydantic import BaseModel, Field
|
||||
@@ -14,6 +15,7 @@ from pydantic import BaseModel, Field
|
||||
from app.application.chat import ChatUseCase
|
||||
from app.application.generate_page import GeneratePageUseCase
|
||||
from app.core.config import Settings, get_settings
|
||||
from app.core.settings_store import save_overrides
|
||||
from app.domain.models import (
|
||||
ArcSummary,
|
||||
CampaignStructuralContext,
|
||||
@@ -29,6 +31,7 @@ from app.domain.models import (
|
||||
)
|
||||
from app.domain.ports import LLMProvider, LLMProviderError
|
||||
from app.infrastructure.ollama_adapter import OllamaLLMProvider
|
||||
from app.infrastructure.onemin_adapter import OneMinAiLLMProvider
|
||||
|
||||
app = FastAPI(
|
||||
title="LoreMind Brain",
|
||||
@@ -189,10 +192,17 @@ def get_llm_provider(
|
||||
"""Factory d'adapter — point d'inversion de dépendance.
|
||||
|
||||
C'est ici (et uniquement ici) qu'on choisit QUEL adapter concret
|
||||
incarne le port. Pour swap vers un autre fournisseur, on change
|
||||
cette ligne et rien d'autre.
|
||||
incarne le port, en fonction du champ `llm_provider` des Settings
|
||||
(modifiable a chaud depuis l'ecran Parametres de l'UI).
|
||||
"""
|
||||
return OllamaLLMProvider(settings)
|
||||
try:
|
||||
if settings.llm_provider == "onemin":
|
||||
return OneMinAiLLMProvider(settings)
|
||||
return OllamaLLMProvider(settings)
|
||||
except LLMProviderError as exc:
|
||||
# Ex : cle 1min.ai manquante. On renvoie du 400 plutot que du 500
|
||||
# pour que le frontend puisse afficher un message actionnable.
|
||||
raise HTTPException(status_code=400, detail=str(exc)) from exc
|
||||
|
||||
|
||||
def get_generate_page_use_case(
|
||||
@@ -392,6 +402,161 @@ def _to_campaign_context(dto: CampaignContextDTO | None) -> CampaignStructuralCo
|
||||
)
|
||||
|
||||
|
||||
# --- Settings (parametrage runtime depuis l'UI) ------------------------------
|
||||
|
||||
|
||||
class SettingsDTO(BaseModel):
|
||||
"""Vue serialisable des settings modifiables depuis l'UI.
|
||||
|
||||
Expose uniquement les champs que l'utilisateur peut changer a chaud.
|
||||
Les secrets (onemin_api_key) sont masques en lecture.
|
||||
"""
|
||||
|
||||
llm_provider: Literal["ollama", "onemin"]
|
||||
ollama_base_url: str
|
||||
llm_model: str
|
||||
onemin_model: str
|
||||
# True si une cle 1min.ai est deja configuree — pas de leak de la cle elle-meme.
|
||||
onemin_api_key_set: bool
|
||||
|
||||
|
||||
class SettingsUpdateDTO(BaseModel):
|
||||
"""Patch partiel des settings. Tous les champs sont optionnels."""
|
||||
|
||||
llm_provider: Literal["ollama", "onemin"] | None = None
|
||||
ollama_base_url: str | None = None
|
||||
llm_model: str | None = None
|
||||
onemin_model: str | None = None
|
||||
# Chaine vide => on efface la cle. None => pas de changement.
|
||||
onemin_api_key: str | None = None
|
||||
|
||||
|
||||
def _to_settings_dto(s: Settings) -> SettingsDTO:
|
||||
return SettingsDTO(
|
||||
llm_provider=s.llm_provider,
|
||||
ollama_base_url=s.ollama_base_url,
|
||||
llm_model=s.llm_model,
|
||||
onemin_model=s.onemin_model,
|
||||
onemin_api_key_set=bool(s.onemin_api_key),
|
||||
)
|
||||
|
||||
|
||||
@app.get("/settings", response_model=SettingsDTO)
|
||||
def read_settings(settings: Annotated[Settings, Depends(get_settings)]) -> SettingsDTO:
|
||||
"""Retourne la config courante (secrets masques)."""
|
||||
return _to_settings_dto(settings)
|
||||
|
||||
|
||||
@app.put("/settings", response_model=SettingsDTO)
|
||||
def update_settings(patch: SettingsUpdateDTO) -> SettingsDTO:
|
||||
"""Applique un patch partiel aux settings et persiste les overrides.
|
||||
|
||||
Toute requete HTTP suivante verra les nouvelles valeurs (pas de cache).
|
||||
"""
|
||||
overrides = {k: v for k, v in patch.model_dump().items() if v is not None}
|
||||
if overrides:
|
||||
save_overrides(overrides)
|
||||
# Relit .env + overrides fusionnes pour confirmation.
|
||||
return _to_settings_dto(get_settings())
|
||||
|
||||
|
||||
@app.get("/models/ollama")
|
||||
async def list_ollama_models(
|
||||
settings: Annotated[Settings, Depends(get_settings)],
|
||||
) -> dict[str, list[str]]:
|
||||
"""Liste les modeles disponibles sur le serveur Ollama configure.
|
||||
|
||||
Retourne une liste vide si Ollama est injoignable — l'UI affichera un
|
||||
message plutot qu'une 500.
|
||||
"""
|
||||
url = f"{settings.ollama_base_url}/api/tags"
|
||||
try:
|
||||
async with httpx.AsyncClient(timeout=5) as client:
|
||||
response = await client.get(url)
|
||||
response.raise_for_status()
|
||||
data = response.json()
|
||||
except httpx.HTTPError:
|
||||
return {"models": []}
|
||||
models = [m.get("name", "") for m in data.get("models", []) if m.get("name")]
|
||||
return {"models": sorted(models)}
|
||||
|
||||
|
||||
@app.get("/models/onemin")
|
||||
def list_onemin_models() -> dict[str, list[dict[str, object]]]:
|
||||
"""Catalogue statique des modeles 1min.ai, groupes par fournisseur.
|
||||
|
||||
Liste construite par probing direct de l'endpoint chat-with-ai avec
|
||||
une vraie cle API (avril 2026) : chaque ID renvoie 200, les IDs
|
||||
absents renvoient 400 UNSUPPORTED_MODEL.
|
||||
|
||||
Nota : les IDs Anthropic utilisent la nomenclature propre a 1min.ai
|
||||
(`claude-<family>-<version>`), pas la convention officielle Anthropic.
|
||||
"""
|
||||
return {
|
||||
"groups": [
|
||||
{
|
||||
"provider": "Anthropic",
|
||||
"models": ["claude-opus-4-6", "claude-sonnet-4-6"],
|
||||
},
|
||||
{
|
||||
"provider": "OpenAI",
|
||||
"models": [
|
||||
"gpt-5",
|
||||
"gpt-5-mini",
|
||||
"gpt-5-nano",
|
||||
"gpt-4.1",
|
||||
"gpt-4.1-mini",
|
||||
"gpt-4.1-nano",
|
||||
"gpt-4o",
|
||||
"gpt-4o-mini",
|
||||
"gpt-4-turbo",
|
||||
"gpt-3.5-turbo",
|
||||
"o3",
|
||||
"o3-pro",
|
||||
"o3-mini",
|
||||
"o4-mini",
|
||||
],
|
||||
},
|
||||
{
|
||||
"provider": "Google",
|
||||
"models": ["gemini-2.5-pro", "gemini-2.5-flash"],
|
||||
},
|
||||
{
|
||||
"provider": "Mistral",
|
||||
"models": [
|
||||
"mistral-large-latest",
|
||||
"mistral-medium-latest",
|
||||
"mistral-small-latest",
|
||||
"open-mistral-nemo",
|
||||
],
|
||||
},
|
||||
{
|
||||
"provider": "DeepSeek",
|
||||
"models": ["deepseek-chat", "deepseek-reasoner"],
|
||||
},
|
||||
{
|
||||
"provider": "xAI",
|
||||
"models": ["grok-3", "grok-3-mini"],
|
||||
},
|
||||
{
|
||||
"provider": "Meta",
|
||||
"models": [
|
||||
"meta/meta-llama-3.1-405b-instruct",
|
||||
"meta/meta-llama-3-70b-instruct",
|
||||
],
|
||||
},
|
||||
{
|
||||
"provider": "Alibaba",
|
||||
"models": ["qwen-plus", "qwen3-max"],
|
||||
},
|
||||
{
|
||||
"provider": "Perplexity",
|
||||
"models": ["sonar", "sonar-pro"],
|
||||
},
|
||||
]
|
||||
}
|
||||
|
||||
|
||||
def _to_narrative_entity(dto: NarrativeEntityDTO | None) -> NarrativeEntityContext | None:
|
||||
if dto is None:
|
||||
return None
|
||||
|
||||
Reference in New Issue
Block a user