busya/ss-tools
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(agent): proxy gradio config and reduce translate blocking

Browse Source
This commit is contained in:
busya
2026-07-07 17:37:36 +03:00
parent b95df37cd5
commit 0d09e24434
14 changed files with 191 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
agent/src/ss_tools/agent/_config.py
View File

@@ -11,6 +11,7 @@ FASTAPI_URL: str = os.getenv("FASTAPI_URL", "http://localhost:8000")
SERVICE_JWT: str = os.getenv("SERVICE_JWT", "")
GRADIO_SERVER_NAME: str = os.getenv("GRADIO_SERVER_NAME", "0.0.0.0")
GRADIO_SERVER_PORT: int = int(os.getenv("GRADIO_SERVER_PORT", "7860"))
GRADIO_ROOT_PATH: str = os.getenv("GRADIO_ROOT_PATH", "/api/agent/gradio")
GRADIO_ALLOW_PORT_FALLBACK: bool = os.getenv("GRADIO_ALLOW_PORT_FALLBACK", "").strip().lower() in {"1", "true", "yes"}
STORAGE_ROOT: str = os.getenv("STORAGE_ROOT", "/app/storage")
AGENT_PREFETCH_DASHBOARD_LIMIT: int = int(os.getenv("AGENT_PREFETCH_DASHBOARD_LIMIT", "25"))

3
agent/src/ss_tools/agent/app.py
View File

@@ -33,7 +33,7 @@ from langchain_core.messages import HumanMessage
from langchain_openai import ChatOpenAI
from openai import APIConnectionError, APITimeoutError, AuthenticationError, RateLimitError
from ss_tools.agent._config import GRADIO_SERVER_NAME, GRADIO_SERVER_PORT, STORAGE_ROOT as _STORAGE_ROOT
from ss_tools.agent._config import GRADIO_ROOT_PATH, GRADIO_SERVER_NAME, GRADIO_SERVER_PORT, STORAGE_ROOT as _STORAGE_ROOT
from ss_tools.agent._confirmation import (
_pending_confirmations,
confirmation_payload,
@@ -817,5 +817,6 @@ if __name__ == "__main__":
demo.launch(
server_name=GRADIO_SERVER_NAME,
server_port=GRADIO_SERVER_PORT,
root_path=GRADIO_ROOT_PATH,
)
# #endregion AgentChat.GradioApp

10
agent/src/ss_tools/agent/run.py
View File

@@ -12,7 +12,14 @@ import socket
import httpx
from ss_tools.agent._config import FASTAPI_URL, GRADIO_ALLOW_PORT_FALLBACK, GRADIO_SERVER_NAME, GRADIO_SERVER_PORT, SERVICE_JWT
from ss_tools.agent._config import (
FASTAPI_URL,
GRADIO_ALLOW_PORT_FALLBACK,
GRADIO_ROOT_PATH,
GRADIO_SERVER_NAME,
GRADIO_SERVER_PORT,
SERVICE_JWT,
)
from ss_tools.shared.cot_logger import seed_trace_id
from ss_tools.shared.logger import logger
@@ -128,5 +135,6 @@ if __name__ == "__main__":
demo.launch(
server_name=GRADIO_SERVER_NAME,
server_port=port,
root_path=GRADIO_ROOT_PATH,
)
# #endregion AgentChat.Run

2
agent/tests/test_agent/test_run.py
View File

@@ -160,6 +160,7 @@ class TestMainBlock:
patch('ss_tools.agent.langgraph_setup.init_checkpointer'), \
patch('ss_tools.agent.run.SERVICE_JWT', svc_jwt), \
patch('ss_tools.agent.run.GRADIO_SERVER_PORT', gradio_port), \
patch('ss_tools.agent.run.GRADIO_ROOT_PATH', env_overrides.get("GRADIO_ROOT_PATH", "/api/agent/gradio")), \
patch('ss_tools.agent.run.GRADIO_ALLOW_PORT_FALLBACK', gradio_fallback):
mock_asyncio_run.side_effect = lambda coro: coro.close() if hasattr(coro, "close") else None
@@ -206,6 +207,7 @@ class TestMainBlock:
result['set_jwt'].assert_not_called()
result['configure'].assert_not_called()
result['demo'].launch.assert_called_once()
assert result['demo'].launch.call_args.kwargs["root_path"] == "/api/agent/gradio"
def test_main_block_with_service_jwt(self, monkeypatch):
"""Main block sets service JWT via ContextVar."""

10
backend/src/plugins/translate/_batch_proc.py
View File

@@ -13,6 +13,7 @@
# @RATIONALE Extracted from TranslationExecutor. Batch insert delegated to _batch_insert.py.
# @REJECTED Keeping batch processing inside TranslationExecutor — caused class to exceed INV_7.
import asyncio
from datetime import UTC, datetime
import time
from typing import Any
@@ -66,7 +67,7 @@ class BatchProcessingService:
tls = [str(tls)] if not isinstance(tls, list) else tls
# ★ Run local language detection on all rows (heuristic, no LLM)
self._detect_languages(batch_rows, tls)
await self._detect_languages(batch_rows, tls)
source_texts = [r.get("source_text", "") for r in batch_rows if r.get("source_text")]
rc = batch_rows[0].get("source_data") if batch_rows else None
@@ -135,14 +136,15 @@ class BatchProcessingService:
# #region _detect_languages [C:2] [TYPE Function] [SEMANTICS translate,language,detect,lingua]
# @ingroup Translate
# @BRIEF Run local language detection on all batch rows (no LLM).
def _detect_languages(self, batch_rows: list[dict], target_languages: list[str]) -> None:
async def _detect_languages(self, batch_rows: list[dict], target_languages: list[str]) -> None:
"""Run local language detection on all batch rows (no LLM).
Attaches '_detected_lang' (BCP-47 code or 'und') to each row dict.
Uses batch_detect() for efficient multi-text processing.
Uses batch_detect() for efficient multi-text processing, offloaded
to a thread to avoid blocking the event loop.
"""
texts = [row.get("source_text", "") for row in batch_rows]
results = batch_detect(texts, target_languages)
results = await asyncio.to_thread(batch_detect, texts, target_languages)
for row, lang in zip(batch_rows, results):
row["_detected_lang"] = lang
# #endregion _detect_languages

2
backend/src/plugins/translate/executor.py
View File

@@ -18,6 +18,7 @@
# @REJECTED Keeping monolithic executor.py at 1974 lines — violates INV_7 by +1574 lines.
# Single monolithic LLM call — would lose all progress on any failure.
import asyncio
from collections.abc import Callable
from datetime import UTC, datetime
from typing import Any
@@ -145,6 +146,7 @@ class TranslationExecutor:
run.skipped_records = skipped_records
run.cache_hits = cache_hits
self.db.commit()
await asyncio.sleep(0) # yield event loop to prevent starvation (499 on health probes)
batch_id = batch_result.get("batch_id")
if batch_id and batch_result.get("successful", 0) > 0:

25
backend/tests/plugins/translate/test_batch_proc.py
View File

@@ -83,16 +83,37 @@ class TestCreateBatch:
class TestDetectLanguages:
"""Verify _detect_languages."""
def test_detects_languages(self, db_session):
@pytest.mark.asyncio
async def test_detects_languages(self, db_session):
config = MagicMock()
svc = BatchProcessingService(db_session, config)
rows = _batch_rows(2)
with patch('src.plugins.translate._batch_proc.batch_detect',
return_value=["en", "fr"]):
svc._detect_languages(rows, ["fr", "de"])
await svc._detect_languages(rows, ["fr", "de"])
assert rows[0]["_detected_lang"] == "en"
assert rows[1]["_detected_lang"] == "fr"
@pytest.mark.asyncio
async def test_detect_languages_offloads_to_thread(self, db_session):
"""Regression: CPU-bound lingua batch detection must not run on event loop."""
config = MagicMock()
svc = BatchProcessingService(db_session, config)
rows = _batch_rows(2)
async def fake_to_thread(fn, texts, target_languages):
assert fn.__name__ == "batch_detect"
assert texts == ["Hello world 0", "Hello world 1"]
assert target_languages == ["fr"]
return ["en", "en"]
with patch('src.plugins.translate._batch_proc.asyncio.to_thread',
new=AsyncMock(side_effect=fake_to_thread)) as mock_to_thread:
await svc._detect_languages(rows, ["fr"])
mock_to_thread.assert_awaited_once()
assert [row["_detected_lang"] for row in rows] == ["en", "en"]
class TestCheckCache:
"""Verify _check_cache."""

26
backend/tests/plugins/translate/test_executor.py
View File

@@ -230,6 +230,32 @@ class TestProcessBatches:
await executor._process_batches(run, job, batches, ["fr"])
callback.assert_called_once()
@pytest.mark.asyncio
async def test_process_batches_yields_after_commit(self, db_with_run):
"""Regression: batch loop yields to event loop after sync DB commit."""
session, run_id = db_with_run
config = MagicMock()
executor = TranslationExecutor(session, config)
job = _make_job(session)
run = session.query(TranslationRun).filter(TranslationRun.id == run_id).first()
batches = [
[{"row_index": "0", "source_text": "Hello", "source_data": {"table": "t"}}],
[{"row_index": "1", "source_text": "World", "source_data": {"table": "t"}}],
]
with patch.object(executor, '_process_batch', new=AsyncMock(side_effect=[
{"successful": 1, "failed": 0, "skipped": 0, "retries": 0, "cache_hits": 0, "batch_id": "batch-1"},
{"successful": 1, "failed": 0, "skipped": 0, "retries": 0, "cache_hits": 0, "batch_id": "batch-2"},
])):
with patch.object(executor, '_insert_batch_to_target', new=AsyncMock(return_value=None)):
with patch('src.plugins.translate.executor.asyncio.sleep', new=AsyncMock()) as mock_sleep:
result = await executor._process_batches(run, job, batches, ["fr"])
assert result.status == "COMPLETED"
assert mock_sleep.await_count == 2
mock_sleep.assert_any_await(0)
class TestFinalizeRun:
"""Verify _finalize_run."""

70
backend/tests/test_gradio_proxy_config.py Normal file
View File

@@ -0,0 +1,70 @@
# #region Test.Infrastructure.GradioProxy [C:3] [TYPE Module] [SEMANTICS test,nginx,docker,agent,gradio]
# @BRIEF Verify Gradio agent proxy contracts across nginx, compose, and agent launch config.
# @RELATION BINDS_TO -> [docker.nginx.conf]
# @RELATION BINDS_TO -> [docker.nginx.ssl.conf]
# @RELATION BINDS_TO -> [AgentChat.Config]
# @TEST_EDGE: docker_dns_name -> Proxy targets compose service name `agent`.
# @TEST_EDGE: prefix_forwarding -> Proxy preserves `/api/agent/gradio` for Gradio root_path.
# @TEST_EDGE: auth_forwarding -> Browser Authorization header is forwarded.
from pathlib import Path
PROJECT_ROOT = Path(__file__).resolve().parents[2]
def _read(relative_path: str) -> str:
return (PROJECT_ROOT / relative_path).read_text(encoding="utf-8")
def test_nginx_http_gradio_proxy_targets_agent_service_and_preserves_prefix():
"""HTTP nginx routes Gradio traffic to compose service `agent` without stripping root_path."""
text = _read("docker/nginx.conf")
assert "location /api/agent/gradio/" in text
assert "set $agent_api http://agent:7860;" in text
assert "superset-tools-agent" not in text
assert "rewrite ^/api/agent/gradio" not in text
assert "proxy_set_header Authorization $http_authorization;" in text
assert "proxy_set_header Connection $connection_upgrade;" in text
assert "proxy_redirect http://agent:7860/ /api/agent/gradio/;" in text
def test_nginx_ssl_gradio_proxy_matches_http_contract():
"""SSL nginx keeps the same Gradio proxy invariants as HTTP nginx."""
text = _read("docker/nginx.ssl.conf")
assert "map $http_upgrade $connection_upgrade" in text
assert "location /api/agent/gradio/" in text
assert "set $agent_api http://agent:7860;" in text
assert "superset-tools-agent" not in text
assert "rewrite ^/api/agent/gradio" not in text
assert "proxy_set_header Authorization $http_authorization;" in text
assert "proxy_set_header X-Forwarded-Proto https;" in text
def test_compose_frontend_depends_on_agent_and_agent_uses_root_path():
"""Compose starts agent before frontend and configures Gradio root_path."""
for compose_path in ("docker-compose.yml", "docker-compose.enterprise-clean.yml"):
text = _read(compose_path)
assert " frontend:" in text
frontend_block = text.split(" frontend:", 1)[1].split("\n\n", 1)[0]
assert " - backend" in frontend_block
assert " - agent" in frontend_block
agent_block = text.split(" agent:", 1)[1]
assert "GRADIO_ROOT_PATH: /api/agent/gradio" in agent_block
def test_build_script_generated_enterprise_compose_keeps_agent_proxy_contract():
"""Release bundle compose heredocs preserve frontend->agent dependency and root_path."""
text = _read("build.sh")
assert text.count(" - agent") >= 2
assert text.count("GRADIO_ROOT_PATH: /api/agent/gradio") >= 2
def test_agent_launch_uses_default_gradio_root_path():
"""Agent runtime exports `/api/agent/gradio` as Gradio root_path for @gradio/client."""
config_text = _read("agent/src/ss_tools/agent/_config.py")
run_text = _read("agent/src/ss_tools/agent/run.py")
app_text = _read("agent/src/ss_tools/agent/app.py")
assert 'GRADIO_ROOT_PATH: str = os.getenv("GRADIO_ROOT_PATH", "/api/agent/gradio")' in config_text
assert "root_path=GRADIO_ROOT_PATH" in run_text
assert "root_path=GRADIO_ROOT_PATH" in app_text
# #endregion Test.Infrastructure.GradioProxy

4
build.sh
View File

@@ -294,6 +294,7 @@ services:
restart: unless-stopped
depends_on:
- backend
- agent
environment:
SSL_KEY_PASSPHRASE: \${SSL_KEY_PASSPHRASE:-}
ports:
@@ -317,6 +318,7 @@ services:
SERVICE_JWT: \${SERVICE_JWT:?Set SERVICE_JWT in .env}
DATABASE_URL: postgresql+psycopg2://\${POSTGRES_USER:-postgres}:\${POSTGRES_PASSWORD:?Set POSTGRES_PASSWORD}@\${POSTGRES_HOST:?Set POSTGRES_HOST}:\${POSTGRES_PORT:-5432}/\${POSTGRES_DB:-ss_tools}
GRADIO_SERVER_PORT: 7860
GRADIO_ROOT_PATH: /api/agent/gradio
ports:
- "\${AGENT_HOST_PORT:-7860}:7860"
DEPLOY
@@ -610,6 +612,7 @@ services:
restart: unless-stopped
depends_on:
- backend
- agent
environment:
SSL_KEY_PASSPHRASE: \${SSL_KEY_PASSPHRASE:-}
ports:
@@ -633,6 +636,7 @@ services:
SERVICE_JWT: \${SERVICE_JWT:?Set SERVICE_JWT in .env}
DATABASE_URL: postgresql+psycopg2://\${POSTGRES_USER:-postgres}:\${POSTGRES_PASSWORD:?Set POSTGRES_PASSWORD}@\${POSTGRES_HOST:?Set POSTGRES_HOST}:\${POSTGRES_PORT:-5432}/\${POSTGRES_DB:-ss_tools}
GRADIO_SERVER_PORT: 7860
GRADIO_ROOT_PATH: /api/agent/gradio
ports:
- "\${AGENT_HOST_PORT:-7860}:7860"
DEPLOY

2
docker-compose.enterprise-clean.yml
View File

@@ -74,6 +74,7 @@ services:
restart: unless-stopped
depends_on:
- backend
- agent
environment:
# Пароль от приватного ключа/PKCS#12 для SSL терминации nginx.
# Опционально: если server.key не зашифрован или используется HTTP — не требуется.
@@ -105,6 +106,7 @@ services:
SERVICE_JWT: ${SERVICE_JWT:-agent-service-secret}
DATABASE_URL: postgresql+psycopg2://${POSTGRES_USER:-postgres}:${POSTGRES_PASSWORD:?Set POSTGRES_PASSWORD in .env.enterprise-clean}@${POSTGRES_HOST:?Set POSTGRES_HOST in .env.enterprise-clean}:${POSTGRES_PORT:-5432}/${POSTGRES_DB:-ss_tools}
GRADIO_SERVER_PORT: 7860
GRADIO_ROOT_PATH: /api/agent/gradio
LLM_CA_CERT_URLS: ${LLM_CA_CERT_URLS:-}
ports:
- "${AGENT_HOST_PORT:-7860}:7860"

2
docker-compose.yml
View File

@@ -64,6 +64,7 @@ services:
SERVICE_JWT: ${SERVICE_JWT:-agent-service-secret}
DATABASE_URL: postgresql+psycopg2://postgres:postgres@db:5432/ss_tools
GRADIO_SERVER_PORT: 7860
GRADIO_ROOT_PATH: /api/agent/gradio
LLM_CA_CERT_URLS: ${LLM_CA_CERT_URLS:-}
ports:
- "7860" # internal only, proxied through nginx
@@ -77,6 +78,7 @@ services:
restart: unless-stopped
depends_on:
- backend
- agent
ports:
- "${FRONTEND_HOST_PORT:-8000}:80"

21
docker/nginx.conf
View File

@@ -1,3 +1,8 @@
map $http_upgrade $connection_upgrade {
default upgrade;
'' close;
}
server {
listen 80;
server_name _;
@@ -10,6 +15,22 @@ server {
try_files $uri $uri/ /index.html;
}
# Gradio agent proxy — must be before /api/ (longest prefix match wins)
location /api/agent/gradio/ {
set $agent_api http://agent:7860;
proxy_pass $agent_api;
proxy_http_version 1.1;
proxy_set_header Upgrade $http_upgrade;
proxy_set_header Connection $connection_upgrade;
proxy_set_header Host $host;
proxy_set_header Authorization $http_authorization;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto $scheme;
proxy_redirect http://agent:7860/ /api/agent/gradio/;
proxy_read_timeout 300s;
}
location /api/ {
set $backend_api http://backend:8000;
proxy_pass $backend_api;

21
docker/nginx.ssl.conf
View File

@@ -13,6 +13,11 @@
# @REJECTED Прокси на HTTPS внутри compose-сети отвергнут — избыточно.
# #endregion docker.nginx.ssl.conf
map $http_upgrade $connection_upgrade {
default upgrade;
'' close;
}
server {
listen 80 default_server;
server_name _;
@@ -39,6 +44,22 @@ server {
try_files $uri $uri/ /index.html;
}
# Gradio agent proxy — must be before /api/ (longest prefix match wins)
location /api/agent/gradio/ {
set $agent_api http://agent:7860;
proxy_pass $agent_api;
proxy_http_version 1.1;
proxy_set_header Upgrade $http_upgrade;
proxy_set_header Connection $connection_upgrade;
proxy_set_header Host $host;
proxy_set_header Authorization $http_authorization;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header X-Forwarded-Proto https;
proxy_redirect http://agent:7860/ /api/agent/gradio/;
proxy_read_timeout 300s;
}
location /api/ {
set $backend_api http://backend:8000;
proxy_pass $backend_api;