Label API models consistently, widen provider detection with suffix matching, drop tomllib from (#46)

release scripts

Author: piotrekno1

CHANGELOG.md

@@ -7,6 +7,8 @@ under Unreleased and move into a version section at release time.
 
 ### Added
 
+- `register_model()` accepts a `model_format` override for models without a matching extractor; `llm_api()` registers its models as format `"api"`, matching the openai/anthropic integrations instead of `"unknown"`.
+- `source_from_base_url()` recognizes more provider hosts (anthropic, mistral, groq, together, deepseek, xai, google, fireworks, cerebras, perplexity, nvidia, huggingface, baseten) instead of falling back to raw hostnames, including suffix matching for per-resource subdomains (Azure OpenAI, Baseten).
 - `examples/llm_api_example.py`: raw-HTTP LLM tracking with `wildedge.llm_api()`; existing examples updated to the module-level API (`wildedge.span` / `wildedge.flush` / `wildedge.register_model` instead of threading a client variable)
 - Releases ship `llms.txt` and `llms-full.txt` as GitHub release assets: the full documentation for that exact version in one file, generated by `scripts/build_llms_txt.py`. README quickstart rewritten around the module-level API.
 - `wildedge doctor --send-test-event`: sends one real span event through the full pipeline and reports the ingest response, proving DSN auth and connectivity end to end. The report gains an `environment` section (`WILDEDGE_*` variables, autoload PYTHONPATH status) plus `config_status` / `connectivity_status` fields.

docs/manual-tracking.md

@@ -81,11 +81,12 @@ For remote APIs with no local object to inspect, pass a placeholder:
 
 ```python
 handle = client.register_model(
-    object(),
+    None,
     model_id="openai/gpt-4o",
-    source="https://api.openai.com",
+    source="openai",
     family="gpt-4o",
     version="2024-08-06",
+    model_format="api",
 )
 ```
 

scripts/check_tag_version.py

@@ -4,19 +4,23 @@
 from __future__ import annotations
 
 import os
+import re
 from pathlib import Path
 
-import tomllib
-
 
 def main() -> None:
     tag = os.environ["TAG_NAME"]
     if not tag.startswith("v"):
         raise SystemExit(f"Expected tag to start with 'v', got: {tag}")
     tag_version = tag[1:]
 
-    data = tomllib.loads(Path("pyproject.toml").read_text())
-    project_version = data["project"]["version"]
+    # Regex instead of tomllib so the script also runs on Python 3.10.
+    match = re.search(
+        r'^version = "([^"]+)"', Path("pyproject.toml").read_text(), re.MULTILINE
+    )
+    if match is None:
+        raise SystemExit("could not find project version in pyproject.toml")
+    project_version = match.group(1)
 
     if tag_version != project_version:
         raise SystemExit(

tests/test_client.py

@@ -201,3 +201,31 @@ def test_app_identity_env_override_used_for_paths(monkeypatch):
     p.assert_called_once_with("env-app")
     d.assert_called_once_with("env-app")
     r.assert_called_once_with("env-app")
+
+
+def test_register_model_defaults_name_to_model_id_and_accepts_format():
+    from wildedge.client import WildEdge
+
+    client = WildEdge(dsn="https://secret@ingest.wildedge.dev/key")
+    client.registry.register.return_value = (object(), True)
+
+    client.register_model(
+        None, model_id="org/model", source="openrouter", model_format="api"
+    )
+
+    info = client.registry.register.call_args[0][1]
+    assert info.model_name == "org/model"
+    assert info.model_source == "openrouter"
+    assert info.model_format == "api"
+
+
+def test_register_model_format_defaults_to_unknown():
+    from wildedge.client import WildEdge
+
+    client = WildEdge(dsn="https://secret@ingest.wildedge.dev/key")
+    client.registry.register.return_value = (object(), True)
+
+    client.register_model(None, model_id="org/model")
+
+    info = client.registry.register.call_args[0][1]
+    assert info.model_format == "unknown"

tests/test_integrations_openai.py

@@ -183,7 +183,19 @@ def register_model(obj, *, model_id=None, source=None, **kwargs):
         ("", "openai"),
         ("https://openrouter.ai/api/v1", "openrouter"),
         ("https://api.openai.com/v1", "openai"),
-        ("https://api.together.xyz/v1", "api.together.xyz"),
+        ("https://api.together.xyz/v1", "together"),
+        ("https://api.anthropic.com", "anthropic"),
+        ("https://api.mistral.ai/v1", "mistral"),
+        ("https://api.groq.com/openai/v1", "groq"),
+        ("https://generativelanguage.googleapis.com/v1beta/openai", "google"),
+        ("https://api.fireworks.ai/inference/v1", "fireworks"),
+        ("https://myresource.openai.azure.com/openai/v1", "azure-openai"),
+        (
+            "https://model-abc123.api.baseten.co/environments/production/sync/v1",
+            "baseten",
+        ),
+        ("https://inference.baseten.co/v1", "baseten"),
+        ("https://unknown.example.com/v1", "unknown.example.com"),
         ("https://localhost:11434/v1", "localhost"),
     ],
 )

tests/test_llm_api.py

@@ -221,3 +221,10 @@ def test_noop_without_dsn_end_to_end(monkeypatch):
         call.usage(tokens_in=1, tokens_out=2)
 
     assert wildedge.get_client().noop is True
+
+
+def test_registers_model_as_api_format(client):
+    with wildedge.llm_api(model="m", provider="p"):
+        pass
+
+    assert client.registered[0]["model_format"] == "api"

wildedge/client.py

@@ -502,13 +502,15 @@ def register_model(
         family: str | None = None,
         version: str | None = None,
         quantization: str | None = None,
+        model_format: str | None = None,
         auto_instrument: bool = True,
     ) -> ModelHandle:
         """
         Register a model and return a handle for tracking events.
 
         Auto-extracts metadata from ONNX Runtime and GGUF/llama.cpp objects.
-        User-supplied kwargs override extracted values.
+        User-supplied kwargs override extracted values. ``model_format``
+        applies when no extractor matches (e.g. ``"api"`` for remote models).
         """
         overrides = {
             k: v
@@ -518,6 +520,7 @@ def register_model(
                 "family": family,
                 "version": version,
                 "quantization": quantization,
+                "format": model_format,
             }.items()
             if v is not None
         }

wildedge/integrations/common.py

@@ -26,12 +26,39 @@ def debug_failure(framework: str, context: str, exc: BaseException) -> None:
 SOURCE_BY_HOSTNAME: dict[str, str] = {
     "api.openai.com": "openai",
     "openrouter.ai": "openrouter",
+    "api.anthropic.com": "anthropic",
+    "api.mistral.ai": "mistral",
+    "api.groq.com": "groq",
+    "api.together.xyz": "together",
+    "api.deepseek.com": "deepseek",
+    "api.x.ai": "xai",
+    "generativelanguage.googleapis.com": "google",
+    "api.fireworks.ai": "fireworks",
+    "api.cerebras.ai": "cerebras",
+    "api.perplexity.ai": "perplexity",
+    "integrate.api.nvidia.com": "nvidia",
+    "router.huggingface.co": "huggingface",
+    "inference.baseten.co": "baseten",
 }
 
+# Providers whose endpoints live on per-resource subdomains.
+SOURCE_BY_HOSTNAME_SUFFIX: list[tuple[str, str]] = [
+    (".openai.azure.com", "azure-openai"),
+    (".api.baseten.co", "baseten"),
+]
+
 
 def source_from_base_url(base_url: str | None) -> str:
     hostname = urlparse(base_url.lower()).hostname if base_url else ""
-    return SOURCE_BY_HOSTNAME.get(hostname or "", hostname or "openai")
+    if not hostname:
+        return "openai"
+    exact = SOURCE_BY_HOSTNAME.get(hostname)
+    if exact is not None:
+        return exact
+    for suffix, source in SOURCE_BY_HOSTNAME_SUFFIX:
+        if hostname.endswith(suffix):
+            return source
+    return hostname
 
 
 def _msg_role(m) -> str | None:

wildedge/llm_api.py

@@ -179,7 +179,7 @@ def _handle(self) -> ModelHandle | None:
 
         try:
             return get_client().register_model(
-                None, model_id=self.model, source=self.source
+                None, model_id=self.model, source=self.source, model_format="api"
             )
         except Exception as exc:
             logger.debug("wildedge: llm model registration failed: %s", exc)