Add AI-augmented codegen: openapi, tests, register, config-docs subco… by wrhalpin · Pull Request #98 · wrhalpin/GNAT

wrhalpin · 2026-04-09T13:34:32Z

…mmands

gnat/codegen/openapi_generator.py: add --ai/--config flags; _try_load_llm(), _ai_enhance(), _ai_test_fixtures() helpers; ai_impls injection in _render_client() and _render_tests() so Claude generates complete implementations instead of scaffolds
gnat/codegen/test_generator.py (new): generate_connector_tests() introspects any CLIENT_REGISTRY connector and produces a full pytest scaffold; --ai generates realistic fixtures via Claude
gnat/codegen/registry_sync.py (new): scan_unregistered() finds connector client.py files missing from CLIENT_REGISTRY; sync_registry() patches gnat/clients/init.py in-place with sorted import + registry entry
gnat/codegen/config_docs_generator.py (new): generate_config_docs() reads config.ini.example and splices a per-section Markdown table into docs/reference/configuration.md between sentinel comments; --ai enriches descriptions
gnat/codegen/init.py: export all four new public functions
gnat/cli/main.py: wire --ai/--config onto openapi subparser; add tests, register, and config-docs subparsers with full argument definitions and _cmd_codegen() handlers

https://claude.ai/code/session_01BDoue9HxB83ijLzFARAugq

…mmands - gnat/codegen/openapi_generator.py: add --ai/--config flags; _try_load_llm(), _ai_enhance(), _ai_test_fixtures() helpers; ai_impls injection in _render_client() and _render_tests() so Claude generates complete implementations instead of scaffolds - gnat/codegen/test_generator.py (new): generate_connector_tests() introspects any CLIENT_REGISTRY connector and produces a full pytest scaffold; --ai generates realistic fixtures via Claude - gnat/codegen/registry_sync.py (new): scan_unregistered() finds connector client.py files missing from CLIENT_REGISTRY; sync_registry() patches gnat/clients/__init__.py in-place with sorted import + registry entry - gnat/codegen/config_docs_generator.py (new): generate_config_docs() reads config.ini.example and splices a per-section Markdown table into docs/reference/configuration.md between sentinel comments; --ai enriches descriptions - gnat/codegen/__init__.py: export all four new public functions - gnat/cli/main.py: wire --ai/--config onto openapi subparser; add tests, register, and config-docs subparsers with full argument definitions and _cmd_codegen() handlers https://claude.ai/code/session_01BDoue9HxB83ijLzFARAugq

Copilot

Pull request overview

Adds new AI-augmented codegen utilities to GNAT (OpenAPI connector generation, unit test scaffolding, CLIENT_REGISTRY syncing, and config-docs regeneration) and wires them into the gnat codegen CLI.

Changes:

Extend OpenAPI connector generator with --ai/--config to optionally produce AI-complete implementations + richer fixtures.
Add new codegen modules: connector test generator, CLIENT_REGISTRY sync tool, and config-docs generator (INI → docs injection).
Expose new codegen APIs from gnat.codegen and add new gnat codegen tests/register/config-docs subcommands.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
gnat/codegen/test_generator.py	New generator that introspects `CLIENT_REGISTRY` connectors and emits pytest scaffolds (optionally AI-enhanced fixtures).
gnat/codegen/registry_sync.py	New helper to scan `gnat/connectors/*/client.py` and patch missing imports/registry entries into `gnat/clients/__init__.py`.
gnat/codegen/openapi_generator.py	Adds AI enhancement + fixture generation and new CLI flags, and injects AI-generated method bodies into rendered output.
gnat/codegen/config_docs_generator.py	New generator that parses `config/config.ini.example` and splices per-connector markdown tables into docs (optionally AI-enriched descriptions).
gnat/codegen/init.py	Exports the new codegen public functions.
gnat/cli/main.py	Adds CLI plumbing for `--ai/--config` and new `tests`, `register`, and `config-docs` codegen subcommands.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-09T13:42:28Z

+                monkeypatch.setattr(client, "put", mock_put)
+                monkeypatch.setattr(client, "post", MagicMock())
+                client.upsert_object("{primary_stix_type}", {{"id": "obj-1", "name": "test"}})
+                assert mock_put.called or True  # some connectors may POST for updates


In the generated scaffold, assert mock_put.called or True will always pass, so this test provides no signal. Consider asserting the expected update verb (PUT) for connectors that support it, or parameterize the generator to emit an assertion that actually fails when the wrong method is used.

Suggested change

monkeypatch.setattr(client, "put", mock_put)

monkeypatch.setattr(client, "post", MagicMock())

client.upsert_object("{primary_stix_type}", {{"id": "obj-1", "name": "test"}})

assert mock_put.called or True # some connectors may POST for updates

mock_post = MagicMock(return_value={native_fixture})

monkeypatch.setattr(client, "put", mock_put)

monkeypatch.setattr(client, "post", mock_post)

client.upsert_object("{primary_stix_type}", {{"id": "obj-1", "name": "test"}})

assert mock_put.called or mock_post.called, (

"Expected update to use PUT or POST"

)

Copilot · 2026-04-09T13:42:28Z

+        class TestAuthentication:
+            def test_auth_headers_populated(self, monkeypatch):
+                """authenticate() must set at least one header."""
+                c = {class_name}({auth_kwargs_str})
+                # Patch the underlying _request so no network call is made
+                monkeypatch.setattr(c, "_request", MagicMock(
+                    return_value=type("R", (), {{"status": 200, "data": b\'{{"access_token":"tok"}}\'}})()
+                ))
+                try:
+                    c.authenticate()
+                except Exception:
+                    pass  # Some auth flows need live creds; just ensure no crash on import
+                # The class must have _auth_headers as a dict
+                assert isinstance(c._auth_headers, dict)


The generated auth test mocks BaseClient._request to return an object with .status/.data, but BaseClient._request() actually returns parsed JSON (dict/list/str/None). With the current mock and broad except Exception: pass, the test can pass even if authenticate() never populates headers (it only checks _auth_headers is a dict, which is already true from BaseClient.init). Consider mocking _request to return a dict like {..."access_token": "tok"...} and asserting _auth_headers becomes non-empty (or that a specific header key is set) without swallowing all exceptions.

Copilot · 2026-04-09T13:42:28Z

+    # --- Optional AI fixtures ---
+    ai_fixtures: dict[str, str] = {}
+    if use_ai:
+        from gnat.codegen.openapi_generator import _try_load_llm
+        llm = _try_load_llm(config_path)
+        if llm is None:


generate_connector_tests() imports _try_load_llm from openapi_generator, but _try_load_llm is a private helper (leading underscore) and is also being reused by other modules. This creates tight coupling between unrelated generators and forces type: ignore elsewhere. Consider moving LLM-loading into a small shared helper module (e.g. gnat.codegen.ai_utils) or making it a supported public function.

Copilot · 2026-04-09T13:42:29Z

+    # Optional AI enhancement for descriptions
+    llm = None
+    if use_ai:
+        from gnat.codegen.openapi_generator import _try_load_llm  # type: ignore[attr-defined]
+        llm = _try_load_llm(config_path)
+        if llm is None:


generate_config_docs() imports _try_load_llm from openapi_generator (and suppresses typing with type: ignore). Since _try_load_llm is a private helper, this cross-module dependency is brittle. Consider moving the LLM-loading helper into a shared, non-private location (or exposing it as a public API) so codegen modules don't depend on each other’s internals.

Copilot · 2026-04-09T13:42:29Z

+def _splice(source: str, key: str, content: str) -> str:
+    """
+    Replace the text between sentinel comments in *source* with *content*.
+
+    If no sentinels are found, appends a new block at the end (before the
+    final ``---`` separator if present).
+    """
+    begin_tag = f"<!-- codegen:begin:{key} -->"
+    end_tag   = f"<!-- codegen:end:{key} -->"
+
+    if begin_tag in source and end_tag in source:
+        pattern = re.compile(
+            re.escape(begin_tag) + r".*?" + re.escape(end_tag),
+            re.DOTALL,
+        )
+        replacement = f"{begin_tag}\n{content}\n{end_tag}"
+        return pattern.sub(replacement, source)
+
+    # No sentinel found — inject before the last "---\n" or at end
+    block = f"\n{begin_tag}\n{content}\n{end_tag}\n"
+    last_sep = source.rfind("\n---\n")
+    if last_sep != -1:
+        return source[:last_sep] + block + source[last_sep:]
+    return source + block


The default target docs file (docs/reference/configuration.md) currently does not contain the  /  sentinels. In that case _splice() will append the generated connector block near the end of the file, which likely isn’t the intended location (the docs already have a “### Platform sections” area). Consider adding the sentinel comments to the docs file in this PR, or enhancing _splice() to insert at a deterministic spot (e.g., immediately after the “### Platform sections” heading) when sentinels are absent.

Copilot · 2026-04-09T13:42:29Z

+# Regex to detect the class definition in client.py
+_CLASS_RE = re.compile(
+    r"^class\s+(\w+Client)\s*\(", re.MULTILINE


_CLASS_RE only matches class definitions of the form class <SomethingClient>(...), but the docstring for _detect_class_name() says it finds a BaseClient subclass. As written, it will miss valid connectors where the main class doesn’t end with Client or uses class X: without parentheses, and it may match the wrong class name if multiple classes end with Client. Consider detecting the BaseClient subclass more robustly (e.g., parse the AST and find classes inheriting BaseClient, or at least relax the regex and validate the base list contains BaseClient).

Suggested change

# Regex to detect the class definition in client.py

_CLASS_RE = re.compile(

r"^class\s+(\w+Client)\s*\(", re.MULTILINE

# Regex to detect the BaseClient subclass defined in client.py

_CLASS_RE = re.compile(

r"^class\s+([A-Za-z_]\w*)\s*\(\s*[^)]*\bBaseClient\b[^)]*\)\s*:",

re.MULTILINE,

Copilot · 2026-04-09T13:42:29Z

+def _insert_import(source: str, import_line: str) -> str:
+    """Insert an import line into the sorted block of connector imports."""
+    lines = source.splitlines(keepends=True)
+
+    # Find all existing 'from gnat.connectors.' import lines
+    connector_import_re = re.compile(r"^from gnat\.connectors\.")
+    indices = [i for i, ln in enumerate(lines) if connector_import_re.match(ln)]
+
+    if not indices:
+        # Fallback: insert after the last 'from gnat.' import
+        from_gnat_re = re.compile(r"^from gnat\.")
+        indices = [i for i, ln in enumerate(lines) if from_gnat_re.match(ln)]
+        if not indices:
+            return source
+
+    # Insert in alphabetical order within the connector import block
+    insert_line = import_line + "\n"
+
+    # Find the right position (keep sorted)
+    for idx in indices:
+        if lines[idx].strip() >= import_line:
+            lines.insert(idx, insert_line)
+            return "".join(lines)
+
+    # Append after the last connector import
+    last_idx = indices[-1]
+    lines.insert(last_idx + 1, insert_line)
+    return "".join(lines)


_insert_import() unconditionally inserts import_line and doesn’t check whether the exact import already exists. If gnat/clients/__init__.py already has the import (but the registry key is missing), sync_registry() can introduce duplicate imports. Consider adding a guard like if import_line in source: return source before attempting insertion.

Copilot · 2026-04-09T13:42:30Z

+def _try_load_llm(config_path: str | None) -> Any:
+    """
+    Attempt to load LLMClient from GNAT config.
+
+    Returns ``None`` if Claude is not configured or dependencies are missing.
+    """
+    try:
+        from gnat.config import GNATConfig
+        from gnat.agents.llm import LLMClient
+
+        cfg = GNATConfig(config_path=config_path)
+        claude_cfg = cfg.get("claude")
+        return LLMClient(backend="claude", **claude_cfg)
+    except Exception:
+        return None
+


_try_load_llm() is defined as a private helper but is imported and reused by other codegen modules (tests/config-docs). Consider moving this into a shared utility module or making it a supported public function, so these modules don’t depend on openapi_generator internals.

Copilot AI review requested due to automatic review settings April 9, 2026 13:34

wrhalpin merged commit 14bc66f into main Apr 9, 2026
1 of 7 checks passed

Copilot started reviewing on behalf of wrhalpin April 9, 2026 13:35 View session

Copilot AI reviewed Apr 9, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add AI-augmented codegen: openapi, tests, register, config-docs subco…#98

Add AI-augmented codegen: openapi, tests, register, config-docs subco…#98
wrhalpin merged 1 commit intomainfrom
claude/add-claude-documentation-k8vvJ

wrhalpin commented Apr 9, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Copilot AI Apr 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

wrhalpin commented Apr 9, 2026

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants