mrleo1nid/gpu-rent
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add opt-in /assistent/chat-eval for AssistentChat via Debug API.

Browse Source 
Lets agents POST/GET a real Assistent turn (tunnel or SSH) without folding it into cheap /snapshot; documents VRAM/Sqlite side effects.

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
Leonid Pershin
2026-08-23 07:21:13 +03:00
co-authored by Cursor
parent 23b07672d6
commit 719d77efd9
4 changed files with 766 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/gpu_rent/debug_api.py
+163 -13
View File
@@ -1,6 +1,8 @@
"""Localhost read-only debug HTTP sidecar for gpu-rent.
"""Localhost debug HTTP sidecar for gpu-rent.
Bind only 127.0.0.1. Started from `up` / `tunnel` / `debug`. No mutations.
Bind only 127.0.0.1. Started from `up` / `tunnel` / `debug`.
Mostly read-only; `/assistent/chat-eval` is an opt-in AssistentChat probe
(may load VRAM / persist chat) and is not part of `/snapshot`.
"""
from __future__ import annotations
@@ -28,14 +30,49 @@ CACHE_TTL_DIAG = 30.0
CACHE_TTL_LOGS = 10.0
CACHE_TTL_DEFAULT = 8.0
_CHAT_EVAL_PARAMS = [
{
"name": "message",
"in": "query",
"schema": {"type": "string"},
"description": "User turn (default: short RU checkpoint/steps/cfg ask)",
},
{
"name": "persona",
"in": "query",
"schema": {"type": "string"},
"description": "Persona id (default neutral)",
},
{
"name": "pack",
"in": "query",
"schema": {"type": "string"},
"description": "Pack id (default ordinary)",
},
{
"name": "model",
"in": "query",
"schema": {"type": "string"},
"description": "Ollama chat model override (else preferred / default_chat)",
},
{
"name": "timeout",
"in": "query",
"schema": {"type": "number", "default": 120},
"description": "Seconds, capped (15–300)",
},
]
_OPENAPI: dict[str, Any] = {
"openapi": "3.0.3",
"info": {
"title": "gpu-rent Debug API",
"version": "1.0.0",
"version": "1.1.0",
"description": (
"Read-only localhost diagnostics for installer progress and "
"SwarmUI / Comfy / Ollama / Assistent. No mutations."
"Localhost diagnostics for installer progress and "
"SwarmUI / Comfy / Ollama / Assistent. Mostly read-only; "
"POST/GET /assistent/chat-eval is opt-in AssistentChat "
"(may load VRAM / write chat) and is skipped from /snapshot."
),
},
"servers": [{"url": "http://127.0.0.1:17821"}],
@@ -57,7 +94,9 @@ _OPENAPI: dict[str, Any] = {
],
}
},
"/snapshot": {"get": {"summary": "Cheap aggregate (no full diag)"}},
"/snapshot": {
"get": {"summary": "Cheap aggregate (no full diag, no chat-eval)"}
},
"/status": {"get": {"summary": "JSON mirror of gpu-rent status"}},
"/state": {"get": {"summary": "state.json without secrets"}},
"/config": {"get": {"summary": "Config with tokens/passwords redacted"}},
@@ -132,6 +171,38 @@ _OPENAPI: dict[str, Any] = {
"/assistent/logs": {
"get": {"summary": "journalctl swarmui filtered for Assistent/Sqlite/build"}
},
"/assistent/chat-eval": {
"get": {
"summary": (
"Opt-in AssistentChat eval (GetNewSession→AssistentChat); "
"may load VRAM / write chat — not in /snapshot"
),
"parameters": _CHAT_EVAL_PARAMS,
},
"post": {
"summary": (
"Same as GET; prefer JSON body "
"{message,persona,pack,model,timeout}"
),
"requestBody": {
"required": False,
"content": {
"application/json": {
"schema": {
"type": "object",
"properties": {
"message": {"type": "string"},
"persona": {"type": "string"},
"pack": {"type": "string"},
"model": {"type": "string"},
"timeout": {"type": "number"},
},
}
}
},
},
},
},
},
}
@@ -306,27 +377,51 @@ def _make_handler(hub: DebugHub):
)
self._send(code, body, "application/json; charset=utf-8")
def _path_summary(self, path: str) -> str:
ops = _OPENAPI["paths"].get(path) or {}
for method in ("get", "post"):
if method in ops and isinstance(ops[method], dict):
return str(ops[method].get("summary") or method)
return ""
def _html_index(self) -> None:
links = sorted(p for p in _OPENAPI["paths"] if p != "/")
rows = "\n".join(
f'<li><a href="{p}">{p}</a> — '
f'{_OPENAPI["paths"][p]["get"]["summary"]}</li>'
f'<li><a href="{p}">{p}</a> — {self._path_summary(p)}</li>'
for p in links
)
html = (
"<!DOCTYPE html><html><head><meta charset=utf-8>"
"<title>gpu-rent debug</title></head><body>"
f"<h1>gpu-rent Debug API</h1>"
f"<p>base <code>{hub.base_url}</code> · read-only · 127.0.0.1</p>"
f"<p>base <code>{hub.base_url}</code> · mostly read-only · 127.0.0.1</p>"
f"<p>Agent: start at <a href='/openapi.json'>/openapi.json</a> "
f"then <a href='/snapshot'>/snapshot</a>.</p>"
f"then <a href='/snapshot'>/snapshot</a>. "
f"Opt-in chat: <code>/assistent/chat-eval</code> (not in snapshot).</p>"
f"<ul>{rows}</ul></body></html>"
)
self._send(200, html.encode("utf-8"), "text/html; charset=utf-8")
def _read_json_body(self) -> dict[str, Any]:
length = int(self.headers.get("Content-Length") or 0)
if length <= 0:
return {}
raw = self.rfile.read(min(length, 1_000_000))
if not raw:
return {}
try:
data = json.loads(raw.decode("utf-8", "replace"))
except json.JSONDecodeError as exc:
raise ValueError(f"invalid JSON body: {exc}") from exc
if data is None:
return {}
if not isinstance(data, dict):
raise ValueError("JSON body must be an object")
return data
def do_GET(self) -> None: # noqa: N802
try:
self._dispatch()
self._dispatch("GET")
except Exception as exc:
self._json(
500,
@@ -337,19 +432,47 @@ def _make_handler(hub: DebugHub):
},
)
def _dispatch(self) -> None:
def do_POST(self) -> None: # noqa: N802
try:
self._dispatch("POST")
except Exception as exc:
self._json(
500,
{
"ok": False,
"error": str(exc)[:300],
"trace": traceback.format_exc()[-800:],
},
)
def _dispatch(self, method: str = "GET") -> None:
parsed = urlparse(self.path)
path = parsed.path.rstrip("/") or "/"
qs = parse_qs(parsed.query)
if path == "/":
if method != "GET":
self._json(405, {"ok": False, "error": "GET only"})
return
self._html_index()
return
if path == "/openapi.json":
if method != "GET":
self._json(405, {"ok": False, "error": "GET only"})
return
doc = dict(_OPENAPI)
doc["servers"] = [{"url": hub.base_url}]
self._json(200, doc)
return
if method == "POST" and path != "/assistent/chat-eval":
self._json(
405,
{
"ok": False,
"error": "POST only allowed for /assistent/chat-eval",
},
)
return
if path == "/progress":
self._json(200, hub.progress())
return
@@ -534,6 +657,30 @@ def _make_handler(hub: DebugHub):
CACHE_TTL_LOGS,
lambda: debug_assistent.collect_assistent_logs(hub.cfg),
)
elif sub == "chat-eval":
body: dict[str, Any] = {}
if method == "POST":
try:
body = self._read_json_body()
except ValueError as exc:
self._json(400, {"ok": False, "error": str(exc)})
return
def _q(name: str) -> str | None:
if name in body and body[name] is not None:
return body[name]
vals = qs.get(name)
return vals[0] if vals else None
# Never cache — live AssistentChat / VRAM side effects.
payload = debug_assistent.run_assistent_chat_eval(
hub.cfg,
message=_q("message"),
persona=_q("persona"),
pack=_q("pack"),
model=_q("model"),
timeout=_q("timeout"),
)
else:
self._json(
404,
@@ -549,6 +696,7 @@ def _make_handler(hub: DebugHub):
"/assistent/api",
"/assistent/wanted",
"/assistent/logs",
"/assistent/chat-eval",
],
},
)
@@ -611,6 +759,7 @@ def start_debug_server(
f" {base}/openapi.json",
f" {base}/snapshot",
f" {base}/assistent",
f" {base}/assistent/chat-eval (opt-in; not in snapshot)",
]
if log:
for line in lines:
@@ -646,7 +795,8 @@ def run_debug_blocking(
server.set_step("idle_debug")
try:
if log:
log("Debug API слушает (Ctrl+C — выход). Мутаций нет.")
log("Debug API слушает (Ctrl+C — выход). "
"Мутаций конфига/GPU нет; /assistent/chat-eval — opt-in AssistentChat.")
while True:
time.sleep(3600)
except KeyboardInterrupt: