proxyagent 0.8.0__tar.gz → 0.9.0__tar.gz

{proxyagent-0.8.0 → proxyagent-0.9.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: proxyagent
-Version: 0.8.0
+Version: 0.9.0
 Summary: Run any agent (Claude, Codex, custom) on any machine — with no API key on the machine. A secure, self-hosted proxy for models and tools.
 Project-URL: Homepage, https://github.com/teddyoweh/proxyagent
 Author-email: Spawn Labs <teddy@spawnlabs.ai>
@@ -179,9 +179,9 @@ is to centralise *all* of them so the machine running the harness holds only a `
 | **Gemini CLI** | Google | API key · OAuth · Vertex |
 
 Connect each mode in the dashboard's **Harnesses** tab (or `proxyagent provider add … --kind`).
-**API key, OAuth, AWS Bedrock, and Azure are wired today** — for Bedrock the proxy holds the AWS
-credentials and **SigV4-signs** the Claude-on-Bedrock request itself, so the machine needs no
-AWS at all. (Vertex lands next.) The model providers below are the *backends* for model-agnostic
+**Every auth mode is wired**: API key, OAuth, AWS Bedrock (the proxy SigV4-signs the Claude-on-Bedrock
+request itself), Azure, and Google Vertex (service-account JSON → access token → Claude-on-Vertex).
+For Bedrock/Vertex the proxy holds the AWS/GCP credentials and signs upstream, so the machine needs no cloud creds at all. The model providers below are the *backends* for model-agnostic
 harnesses (aider, Cline…).
 
 ```bash
@@ -189,6 +189,7 @@ harnesses (aider, Cline…).
 proxyagent provider add anthropic --kind bedrock --key <AWS_SECRET>   # + meta: access_key, region
 proxyagent provider add openai    --kind azure   --key <AZURE_KEY>    # + meta: endpoint
 proxyagent provider add anthropic --kind oauth    --key <OAUTH_TOKEN>
+proxyagent provider add anthropic --kind vertex  --key "$(cat sa.json)"   # + meta: region
 ```
 
 ## Credential pools & failover

{proxyagent-0.8.0 → proxyagent-0.9.0}/README.md

@@ -147,9 +147,9 @@ is to centralise *all* of them so the machine running the harness holds only a `
 | **Gemini CLI** | Google | API key · OAuth · Vertex |
 
 Connect each mode in the dashboard's **Harnesses** tab (or `proxyagent provider add … --kind`).
-**API key, OAuth, AWS Bedrock, and Azure are wired today** — for Bedrock the proxy holds the AWS
-credentials and **SigV4-signs** the Claude-on-Bedrock request itself, so the machine needs no
-AWS at all. (Vertex lands next.) The model providers below are the *backends* for model-agnostic
+**Every auth mode is wired**: API key, OAuth, AWS Bedrock (the proxy SigV4-signs the Claude-on-Bedrock
+request itself), Azure, and Google Vertex (service-account JSON → access token → Claude-on-Vertex).
+For Bedrock/Vertex the proxy holds the AWS/GCP credentials and signs upstream, so the machine needs no cloud creds at all. The model providers below are the *backends* for model-agnostic
 harnesses (aider, Cline…).
 
 ```bash
@@ -157,6 +157,7 @@ harnesses (aider, Cline…).
 proxyagent provider add anthropic --kind bedrock --key <AWS_SECRET>   # + meta: access_key, region
 proxyagent provider add openai    --kind azure   --key <AZURE_KEY>    # + meta: endpoint
 proxyagent provider add anthropic --kind oauth    --key <OAUTH_TOKEN>
+proxyagent provider add anthropic --kind vertex  --key "$(cat sa.json)"   # + meta: region
 ```
 
 ## Credential pools & failover

{proxyagent-0.8.0 → proxyagent-0.9.0}/proxyagent/__init__.py

@@ -16,7 +16,7 @@ from typing import Optional
 
 from .harness import run  # noqa: F401  (the headline SDK call)
 
-__version__ = "0.8.0"
+__version__ = "0.9.0"
 __all__ = ["run", "serve", "create_app", "Config", "Admin", "__version__"]
 
 

{proxyagent-0.8.0 → proxyagent-0.9.0}/proxyagent/config.py

@@ -101,7 +101,7 @@ AUTH_LABELS = {"api_key": "API key", "oauth": "OAuth", "bedrock": "AWS Bedrock",
                "vertex": "Google Vertex", "azure": "Azure"}
 # Auth modes that are fully wired today (just a key swap). Others are surfaced in the
 # UI as "available" and built out (Bedrock SigV4 / Vertex token / OAuth refresh).
-AUTH_READY = {"api_key", "oauth", "bedrock", "azure"}
+AUTH_READY = {"api_key", "oauth", "bedrock", "azure", "vertex"}
 
 
 @dataclass

{proxyagent-0.8.0 → proxyagent-0.9.0}/proxyagent/providers.py

@@ -70,6 +70,11 @@ def build_plans(provider, store: Store | None, body: dict) -> list[tuple]:
                     plans.append(signers.bedrock_plan(c, body))
                 except Exception:  # noqa: BLE001 — skip a malformed bedrock cred
                     pass
+            elif kind == "vertex":
+                try:
+                    plans.append(signers.vertex_plan(c, body))
+                except Exception:  # noqa: BLE001 — skip if SA invalid / token fetch fails
+                    pass
     if provider.key:
         plans.append((provider.endpoint, {**JSON, **provider.auth_headers()}, raw))
     return plans

{proxyagent-0.8.0 → proxyagent-0.9.0}/proxyagent/signers.py

@@ -9,10 +9,12 @@ Vertex (GCP service-account → access token) lands next.
 
 from __future__ import annotations
 
+import base64
 import datetime
 import hashlib
 import hmac
 import json
+import time
 from urllib.parse import quote
 
 
@@ -86,3 +88,67 @@ def bedrock_plan(cred: dict, body: dict):
         access_key=meta.get("access_key", ""), secret_key=cred.get("secret", ""),
         body=raw, session_token=meta.get("session_token"))
     return f"https://{host}{path}", headers, raw
+
+
+# ------------------------------------------------------------------ #
+# Google Vertex AI — service account → OAuth2 access token → Claude-on-Vertex.
+# The proxy holds the service-account key; the machine holds none.
+# ------------------------------------------------------------------ #
+
+def _b64url(b: bytes) -> str:
+    return base64.urlsafe_b64encode(b).rstrip(b"=").decode("ascii")
+
+
+_VERTEX_TOKEN_CACHE: dict[str, tuple[str, int]] = {}   # client_email -> (token, expiry_ms)
+
+
+def vertex_signed_assertion(sa: dict, *, now: int | None = None) -> str:
+    """A signed RS256 JWT bearer assertion for the service account (no network)."""
+    from cryptography.hazmat.primitives import hashes, serialization
+    from cryptography.hazmat.primitives.asymmetric import padding
+    iat = int(now if now is not None else time.time())
+    token_uri = sa.get("token_uri", "https://oauth2.googleapis.com/token")
+    header = {"alg": "RS256", "typ": "JWT"}
+    claims = {"iss": sa["client_email"], "scope": "https://www.googleapis.com/auth/cloud-platform",
+              "aud": token_uri, "iat": iat, "exp": iat + 3600}
+    signing_input = f"{_b64url(json.dumps(header).encode())}.{_b64url(json.dumps(claims).encode())}"
+    key = serialization.load_pem_private_key(sa["private_key"].encode(), password=None)
+    sig = key.sign(signing_input.encode(), padding.PKCS1v15(), hashes.SHA256())
+    return f"{signing_input}.{_b64url(sig)}"
+
+
+def vertex_access_token(sa: dict) -> str:
+    """Exchange the SA assertion for a short-lived access token (cached ~1h)."""
+    import httpx
+    email = sa["client_email"]
+    nowms = int(time.time() * 1000)
+    cached = _VERTEX_TOKEN_CACHE.get(email)
+    if cached and cached[1] - 60_000 > nowms:
+        return cached[0]
+    token_uri = sa.get("token_uri", "https://oauth2.googleapis.com/token")
+    r = httpx.post(token_uri, data={
+        "grant_type": "urn:ietf:params:oauth:grant-type:jwt-bearer",
+        "assertion": vertex_signed_assertion(sa)}, timeout=20)
+    r.raise_for_status()
+    body = r.json()
+    tok, ttl = body["access_token"], body.get("expires_in", 3600)
+    _VERTEX_TOKEN_CACHE[email] = (tok, nowms + ttl * 1000)
+    return tok
+
+
+def vertex_url(project: str, region: str, model: str) -> str:
+    return (f"https://{region}-aiplatform.googleapis.com/v1/projects/{project}"
+            f"/locations/{region}/publishers/anthropic/models/{quote(model, safe='')}:rawPredict")
+
+
+def vertex_plan(cred: dict, body: dict):
+    """(url, headers, body_bytes) for a Claude-on-Vertex rawPredict call."""
+    sa = json.loads(cred["secret"])                      # the full service-account JSON
+    meta = cred.get("meta") or {}
+    region = meta.get("region") or sa.get("region") or "us-east5"
+    project = meta.get("project_id") or sa.get("project_id")
+    payload = {k: v for k, v in body.items() if k != "model"}
+    payload["anthropic_version"] = "vertex-2023-10-16"
+    raw = json.dumps(payload).encode("utf-8")
+    headers = {"content-type": "application/json", "Authorization": f"Bearer {vertex_access_token(sa)}"}
+    return vertex_url(project, region, body.get("model", "")), headers, raw

{proxyagent-0.8.0 → proxyagent-0.9.0}/proxyagent/ui/index.html

@@ -144,7 +144,7 @@ function logo(name,color){color=color||"#888";
 function fieldsFor(uid,kind){
   if(kind==="bedrock")return `<div class="row"><input id="f1_${uid}" placeholder="AWS access key id" style="flex:1"/><input id="f3_${uid}" placeholder="region (us-east-1)" style="width:150px"/></div><input id="f2_${uid}" type="password" placeholder="AWS secret access key" style="width:100%;margin-top:8px"/>`;
   if(kind==="azure")return `<input id="f1_${uid}" placeholder="Azure endpoint URL (…/chat/completions?api-version=…)" style="width:100%"/><input id="f2_${uid}" type="password" placeholder="api-key" style="width:100%;margin-top:8px"/>`;
-  if(kind==="vertex")return `<span class="muted" style="font-size:12px">Vertex (service-account JSON) — coming next.</span>`;
+  if(kind==="vertex")return `<textarea id="f1_${uid}" placeholder="Paste the GCP service-account JSON" style="width:100%;min-height:92px;font:12px ui-monospace,monospace;border:none;border-radius:11px;background:var(--panel2);color:var(--txt);padding:10px 12px;outline:none"></textarea><input id="f3_${uid}" placeholder="region (default us-east5)" style="width:100%;margin-top:8px"/>`;
   return `<input id="f1_${uid}" type="password" placeholder="${kind==='oauth'?'OAuth access token':'API key'}" style="width:100%"/>`;
 }
 function onKind(uid){document.getElementById("ff_"+uid).innerHTML=fieldsFor(uid,document.getElementById("kind_"+uid).value)}
@@ -153,7 +153,7 @@ async function submitCred(uid,provider){
   let secret,meta={};
   if(kind==="bedrock"){secret=g("f2_"+uid);meta={access_key:g("f1_"+uid),region:g("f3_"+uid)||"us-east-1"}}
   else if(kind==="azure"){secret=g("f2_"+uid);meta={endpoint:g("f1_"+uid)}}
-  else if(kind==="vertex"){return}
+  else if(kind==="vertex"){secret=g("f1_"+uid);meta={region:g("f3_"+uid)||"us-east5"}}
   else{secret=g("f1_"+uid)}
   if(!secret)return;
   await api("/admin/providers",{method:"POST",body:JSON.stringify({provider,secret,kind,meta})});refreshKeys();

{proxyagent-0.8.0 → proxyagent-0.9.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "proxyagent"
-version = "0.8.0"
+version = "0.9.0"
 description = "Run any agent (Claude, Codex, custom) on any machine — with no API key on the machine. A secure, self-hosted proxy for models and tools."
 readme = "README.md"
 requires-python = ">=3.10"

{proxyagent-0.8.0 → proxyagent-0.9.0}/tests/test_proxy.py

@@ -236,3 +236,28 @@ def test_bedrock_plan_and_build_plans():
     s.add_credential("anthropic", "awssecret", kind="bedrock", meta={"access_key": "AKID", "region": "us-east-1"})
     plans = build_plans(PROVIDERS["anthropic"], s, {"model": "claude-sonnet-4-5", "messages": []})
     assert len(plans) == 2 and plans[0][0].endswith("/v1/messages") and "bedrock-runtime" in plans[1][0]
+
+
+def test_vertex_assertion_and_url():
+    import base64, json
+    from cryptography.hazmat.primitives import hashes, serialization
+    from cryptography.hazmat.primitives.asymmetric import padding, rsa
+    from proxyagent.signers import vertex_signed_assertion, vertex_url
+    key = rsa.generate_private_key(public_exponent=65537, key_size=2048)
+    pem = key.private_bytes(serialization.Encoding.PEM, serialization.PrivateFormat.PKCS8,
+                            serialization.NoEncryption()).decode()
+    sa = {"client_email": "svc@proj.iam.gserviceaccount.com", "private_key": pem,
+          "token_uri": "https://oauth2.googleapis.com/token"}
+    jwt = vertex_signed_assertion(sa, now=1700000000)
+    parts = jwt.split(".")
+    assert len(parts) == 3
+    b64d = lambda s: base64.urlsafe_b64decode(s + "=" * (-len(s) % 4))
+    header, claims = json.loads(b64d(parts[0])), json.loads(b64d(parts[1]))
+    assert header["alg"] == "RS256" and claims["iss"] == sa["client_email"]
+    assert claims["scope"].endswith("cloud-platform")
+    # signature must verify against the public key (raises on tamper)
+    key.public_key().verify(b64d(parts[2]), (parts[0] + "." + parts[1]).encode(),
+                            padding.PKCS1v15(), hashes.SHA256())
+    assert vertex_url("myproj", "us-east5", "claude-sonnet-4-5") == (
+        "https://us-east5-aiplatform.googleapis.com/v1/projects/myproj/locations/us-east5"
+        "/publishers/anthropic/models/claude-sonnet-4-5:rawPredict")