honour GRAPHIFY_API_TIMEOUT in claude-cli and Anthropic SDK backends (#1112)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

README.md

@@ -366,7 +366,7 @@ These are only needed for **headless / CI extraction** (`graphify extract`). Whe
 | `AWS_*` / `~/.aws/credentials` | AWS Bedrock — standard credential chain | `--backend bedrock` (no API key, uses IAM) |
 | `GRAPHIFY_MAX_WORKERS` | AST parallelism thread count | optional — also `--max-workers` flag |
 | `GRAPHIFY_MAX_OUTPUT_TOKENS` | Raise output cap for dense corpora | optional — e.g. `32768` for large files |
-| `GRAPHIFY_API_TIMEOUT` | HTTP timeout in seconds (default: 600) | optional — also `--api-timeout` flag |
+| `GRAPHIFY_API_TIMEOUT` | Per-call timeout in seconds for HTTP, claude-cli, and Anthropic SDK backends (default: 600) | optional — also `--api-timeout` flag |
 | `GRAPHIFY_FORCE` | Force graph rebuild even with fewer nodes | optional — also `--force` flag |
 | `GRAPHIFY_GOOGLE_WORKSPACE` | Auto-enable Google Workspace export | optional — set to `1` |
 | `GRAPHIFY_TRIAGE_BACKEND` | Backend for `graphify prs --triage` | optional — auto-detected from available keys |

graphify/llm.py

@@ -214,6 +214,19 @@ def _resolve_max_tokens(default: int) -> int:
             pass
     return default
 
+
+def _resolve_api_timeout(default: float = 600.0) -> float:
+    """Honour GRAPHIFY_API_TIMEOUT env var override, else use default (seconds)."""
+    raw = os.environ.get("GRAPHIFY_API_TIMEOUT", "").strip()
+    if raw:
+        try:
+            v = float(raw)
+            if v > 0:
+                return v
+        except ValueError:
+            pass
+    return default
+
 _EXTRACTION_SYSTEM = """\
 You are a graphify semantic extraction agent. Extract a knowledge graph fragment from the files provided.
 Output ONLY valid JSON — no explanation, no markdown fences, no preamble.
@@ -434,16 +447,7 @@ def _call_openai_compat(
     # default. Honour GRAPHIFY_API_TIMEOUT (seconds) for explicit override;
     # default to 600s, which is long enough for a 31B model on a 16k chunk
     # but still bounds runaway connections (issue #792 addendum).
-    timeout_raw = os.environ.get("GRAPHIFY_API_TIMEOUT", "").strip()
+    client = OpenAI(api_key=api_key, base_url=base_url, timeout=_resolve_api_timeout())
-    timeout_s: float = 600.0
-    if timeout_raw:
-        try:
-            v = float(timeout_raw)
-            if v > 0:
-                timeout_s = v
-        except ValueError:
-            pass
-    client = OpenAI(api_key=api_key, base_url=base_url, timeout=timeout_s)
     kwargs: dict = {
         "model": model,
         "messages": [
@@ -550,7 +554,7 @@ def _call_claude(api_key: str, model: str, user_message: str, max_tokens: int =
     except ImportError as exc:
         raise ImportError(_backend_pkg_hint("anthropic", "anthropic")) from exc
 
-    client = anthropic.Anthropic(api_key=api_key)
+    client = anthropic.Anthropic(api_key=api_key, timeout=_resolve_api_timeout())
     resp = client.messages.create(
         model=model,
         max_tokens=max_tokens,
@@ -637,7 +641,7 @@ def _call_claude_cli(user_message: str, max_tokens: int = 8192, *, deep_mode: bo
         capture_output=True,
         text=True,
         encoding="utf-8",  # Force UTF-8 — prevents UnicodeEncodeError on Windows cp1252
-        timeout=600,
+        timeout=_resolve_api_timeout(),
         check=False,
     )
     if proc.returncode != 0:
@@ -1200,7 +1204,7 @@ def _call_llm(prompt: str, *, backend: str, max_tokens: int = 200) -> str:
             capture_output=True,
             text=True,
             encoding="utf-8",  # Force UTF-8 — prevents UnicodeEncodeError on Windows cp1252
-            timeout=600,
+            timeout=_resolve_api_timeout(),
             check=False,
         )
         if proc.returncode != 0:

tests/test_claude_cli_backend.py

@@ -193,3 +193,38 @@ def test_non_windows_uses_bare_claude(monkeypatch):
 
     argv = run.call_args.args[0]
     assert argv[0] == "claude"
+
+
+# ---------- GRAPHIFY_API_TIMEOUT honoured by all backends ----------
+
+
+def test_resolve_api_timeout_default(monkeypatch):
+    monkeypatch.delenv("GRAPHIFY_API_TIMEOUT", raising=False)
+    assert llm._resolve_api_timeout() == 600.0
+
+
+def test_resolve_api_timeout_env_override(monkeypatch):
+    monkeypatch.setenv("GRAPHIFY_API_TIMEOUT", "45")
+    assert llm._resolve_api_timeout() == 45.0
+
+
+def test_resolve_api_timeout_ignores_invalid(monkeypatch):
+    monkeypatch.setenv("GRAPHIFY_API_TIMEOUT", "not-a-number")
+    assert llm._resolve_api_timeout() == 600.0
+
+
+def test_resolve_api_timeout_ignores_nonpositive(monkeypatch):
+    monkeypatch.setenv("GRAPHIFY_API_TIMEOUT", "0")
+    assert llm._resolve_api_timeout() == 600.0
+
+
+def test_claude_cli_extraction_honours_timeout(monkeypatch, fake_claude):
+    monkeypatch.setenv("GRAPHIFY_API_TIMEOUT", "30")
+    llm._call_claude_cli("dummy", max_tokens=8192)
+    assert fake_claude.call_args.kwargs["timeout"] == 30.0
+
+
+def test_call_llm_claude_cli_branch_honours_timeout(monkeypatch, fake_claude):
+    monkeypatch.setenv("GRAPHIFY_API_TIMEOUT", "30")
+    llm._call_llm(prompt="x", backend="claude-cli", max_tokens=10)
+    assert fake_claude.call_args.kwargs["timeout"] == 30.0