Merge pull request #6 from HDMowri/release/v1.0.1

v1.0.1: 18 AI models access with GitHub Copilot SDK + Test infrastructure

Author: mowree

.gitignore

@@ -54,6 +54,11 @@ yarn-error.log*
 pnpm-debug.log*
 lerna-debug.log*
 
+# Dev scratch files
+test_live_registry.py
+analyze_coverage.py
+tests_backup_*/
+
 # Azure
 .azure
 azure.yaml
@@ -76,3 +81,22 @@ next-steps.md
 
 # Test output artifacts
 forensic_report.json
+test_report_*.txt
+test_report_*.json
+test_report_*.html
+e2e_result.txt
+push_result.txt
+push_fork_result.txt
+.pytest_cache_win/
+
+# Test evidence (generated by saturation tests)
+test_evidence/
+
+# Amplifier cache (local)
+.amplifier/
+
+# Internal development artifacts (Windows-specific, not for public repo)
+investigations/
+mydocs
+run_parallel_tests.ps1
+test_results.txt

Makefile

@@ -35,8 +35,9 @@ install:
 
 # Run all tests (uses the 3-command pattern to avoid Windows asyncio issues)
 test:
-	@echo "Running tests (step 1/3: model naming)..."
-	$(PYTHON) -m pytest tests/test_model_naming.py -q --tb=short
+	@echo "Running tests (step 1/3: model naming + cache)..."
+	$(PYTHON) -m pytest tests/test_model_naming.py tests/test_model_cache.py \
+		tests/test_model_cache_integration.py -q --tb=short
 	@echo "Running tests (step 2/3: sync-heavy + SDK assumptions)..."
 	$(PYTHON) -m pytest tests/test_models.py tests/test_converters.py tests/test_client.py \
 		tests/test_exceptions.py tests/test_mount.py tests/test_mount_coverage.py \

README.md

@@ -76,25 +76,15 @@ amplifier provider use github-copilot --local
 amplifier provider use github-copilot --model claude-sonnet-4 --project
 ```
 
-### Provider Preferences in Bundles
+## Supported Models (18)
 
-Use provider preferences for ordered model fallback:
+All 18 models available through your Copilot subscription are exposed at runtime:
 
-```yaml
-provider_preferences:
-  - provider: github-copilot
-    model: claude-sonnet-4
-  - provider: github-copilot
-    model: gpt-*
-```
-
-## Supported Models
+**Anthropic:** `claude-haiku-4.5`, `claude-opus-4.5`, `claude-opus-4.6`, `claude-opus-4.6-1m`, `claude-opus-4.6-fast`, `claude-sonnet-4`, `claude-sonnet-4.5`
 
-All models available through your Copilot subscription are exposed at runtime. Examples:
+**OpenAI:** `gpt-4.1`, `gpt-5`, `gpt-5-mini`, `gpt-5.1`, `gpt-5.1-codex`, `gpt-5.1-codex-max`, `gpt-5.1-codex-mini`, `gpt-5.2`, `gpt-5.2-codex`, `gpt-5.3-codex`
 
-- `claude-opus-4.5`, `claude-sonnet-4`, `claude-haiku-4.5`
-- `gpt-5`, `gpt-5.1`, `gpt-5.1-codex`
-- `gemini-3-pro-preview`
+**Google:** `gemini-3-pro-preview`
 
 ## Configuration
 
@@ -115,7 +105,7 @@ Set `debug: true` for request/response event logging, or `debug: true, raw_debug
 
 ## Contract
 
-| | |
+| Field | Value |
 | --- | --- |
 | **Module Type** | Provider |
 | **Module ID** | `provider-github-copilot` |
@@ -152,6 +142,20 @@ make sdk-assumptions  # Before upgrading SDK
 make check         # Full check (lint + test)
 ```
 
+### Live Integration Tests
+
+Live tests require `RUN_LIVE_TESTS=1` and valid GitHub Copilot authentication:
+
+```bash
+RUN_LIVE_TESTS=1 python -m pytest tests/integration/ -v
+```
+
+On Windows PowerShell:
+
+```powershell
+$env:RUN_LIVE_TESTS="1"; python -m pytest tests/integration/ -v
+```
+
 ## Dependencies
 
 - `amplifier-core` (provided by Amplifier runtime, not installed separately)

amplifier_module_provider_github_copilot/__init__.py

@@ -221,52 +221,26 @@ def _find_copilot_cli(config: dict[str, Any]) -> str | None:
     """
     Find the Copilot CLI executable path.
 
-    Searches in order:
-    1. config["cli_path"] (explicit configuration)
-    2. COPILOT_CLI_PATH environment variable
-    3. System PATH via shutil.which()
+    Uses the SDK's bundled CLI binary by default. Falls back to system PATH
+    if the bundled binary is not found. The SDK bundles its own CLI binary
+    which is version-matched, avoiding potential version mismatches.
 
     Args:
-        config: Provider configuration (read-only)
+        config: Provider configuration (unused, kept for API compatibility)
 
     Returns:
         Resolved CLI path, or None if not found
     """
-    import os
-
     try:
-        # Get CLI path from config or environment
-        cli_path = config.get("cli_path")
-
-        if not cli_path:
-            cli_path = os.environ.get("COPILOT_CLI_PATH")
-
-        # If still not found, try to locate it
-        if not cli_path:
-            found = shutil.which("copilot") or shutil.which("copilot.exe")
-            if found:
-                cli_path = found
-
-        if not cli_path:
-            logger.debug("[MOUNT] Copilot CLI not found in PATH or known locations")
-            return None
-
-        # Verify the path exists
-        if os.path.isabs(cli_path):
-            if not os.path.isfile(cli_path):
-                logger.debug(f"[MOUNT] Copilot CLI not found at: {cli_path}")
-                return None
-            _ensure_executable(cli_path)
-            logger.debug(f"[MOUNT] Found Copilot CLI at absolute path: {cli_path}")
-            return cli_path
-
-        found_path = shutil.which(cli_path)
-        if found_path is None:
-            logger.debug(f"[MOUNT] '{cli_path}' CLI not found in PATH")
-            return None
-        _ensure_executable(found_path)
-        logger.debug(f"[MOUNT] Found Copilot CLI at: {found_path}")
-        return found_path
+        # Let SDK use its bundled CLI - only fall back to system PATH
+        found = shutil.which("copilot") or shutil.which("copilot.exe")
+        if found:
+            _ensure_executable(found)
+            logger.debug(f"[MOUNT] Found Copilot CLI at: {found}")
+            return found
+
+        logger.debug("[MOUNT] Copilot CLI not found in PATH")
+        return None
 
     except Exception as e:
         logger.debug(f"[MOUNT] CLI path discovery failed: {e}")

amplifier_module_provider_github_copilot/_constants.py

@@ -49,6 +49,27 @@
 # Valid reasoning effort levels per Copilot SDK
 VALID_REASONING_EFFORTS = frozenset({"low", "medium", "high", "xhigh"})
 
+# ═══════════════════════════════════════════════════════════════════════════════
+# MODEL CACHE CONFIGURATION
+# ═══════════════════════════════════════════════════════════════════════════════
+#
+# The model cache persists model metadata (context_window, max_output_tokens)
+# to disk so that provider initialization can return accurate values without
+# calling the SDK API every time.
+#
+# Cache is written by list_models() (called during `amplifier init`) and read
+# during provider initialization (mount/constructor).
+#
+# Cross-platform: Uses pathlib.Path.home() which works on:
+# - Linux: /home/<user>/.amplifier/cache/
+# - WSL: /home/<user>/.amplifier/cache/
+# - macOS: /Users/<user>/.amplifier/cache/
+# - Windows: C:\Users\<user>\.amplifier\cache\
+
+CACHE_FORMAT_VERSION = 1  # Increment on breaking schema changes
+CACHE_FILE_NAME = "github-copilot-models.json"  # Provider-specific file
+CACHE_STALE_DAYS = 30  # Log warning if cache older than this
+
 # Maximum repaired tool IDs to track (LRU eviction)
 MAX_REPAIRED_TOOL_IDS = 1000
 
@@ -128,6 +149,20 @@
         # They are NOT documented but conflict with custom tools.
         "report_intent",  # Hidden: Causes hang if custom tool uses same name
         "task",  # Hidden: Causes hang if custom tool uses same name
+        # ─────────────────────────────────────────────────────────────────────────
+        # Additional built-ins discovered via archaeology (2026-02-09) and live
+        # testing (2026-02-16). Bug: GHCP-BUILTIN-TOOLS-001
+        # Evidence: ST04 session, binary analysis, Gemini live test
+        # ─────────────────────────────────────────────────────────────────────────
+        "create",  # File ops: ST04 session - "Tool 'create' not found"
+        "shell",  # Shell: 2026-02-09 archaeology
+        "report_progress",  # Think: 2026-02-09 archaeology (CLI session UI)
+        "update_todo",  # Think: 2026-02-09 archaeology
+        "skill",  # Other: 2026-02-09 archaeology
+        "fetch_copilot_cli_documentation",  # Fetch: 2026-02-16 live Gemini test
+        "search_code_subagent",  # Search: 2026-02-16 binary analysis
+        "github-mcp-server-web_search",  # Search: 2026-02-16 binary analysis (MCP)
+        "task_complete",  # Task: 2026-02-17 forensic session 1541c502
     }
 )
 
@@ -219,4 +254,17 @@ class LoopExitMethod(Enum):
     # These are not documented but cause session hangs if user tool has same name.
     "report_intent": frozenset({"report_intent"}),  # Hidden: Always exclude
     "task": frozenset({"task"}),  # Hidden: Always exclude
+    # ─────────────────────────────────────────────────────────────────────────
+    # Additional built-ins (Bug GHCP-BUILTIN-TOOLS-001, 2026-02-17)
+    # ─────────────────────────────────────────────────────────────────────────
+    "create": frozenset({"write_file"}),  # Maps to write_file (same as edit)
+    "shell": frozenset({"bash"}),  # Maps to bash
+    "update_todo": frozenset({"todo"}),  # Maps to todo
+    "skill": frozenset({"load_skill"}),  # Maps to load_skill
+    "search_code_subagent": frozenset({"grep", "glob", "delegate"}),  # Composite tool
+    "github-mcp-server-web_search": frozenset({"web_search"}),  # MCP web search
+    # Pure exclusions — no direct Amplifier equivalent
+    "report_progress": frozenset({"todo"}),  # Partial: maps to todo for task tracking
+    "fetch_copilot_cli_documentation": frozenset(),  # CLI-specific, no equivalent
+    "task_complete": frozenset({"todo"}),  # Task completion: maps to todo
 }

amplifier_module_provider_github_copilot/client.py

@@ -136,9 +136,9 @@ def __init__(
 
         Args:
             config: Configuration dict with optional keys:
-                - cli_path: Path to Copilot CLI executable
                 - log_level: Logging level for CLI
                 - auto_restart: Whether to auto-restart CLI on crash
+                - cwd: Working directory for CLI process
             timeout: Default request timeout in seconds (must be > 0)
 
         Raises:
@@ -313,23 +313,11 @@ async def ensure_client(self) -> CopilotClient:
     def _build_client_options(self) -> dict[str, Any]:
         """Build CopilotClientOptions from configuration.
 
-        Note: We intentionally do NOT auto-discover copilot CLI from PATH.
         The SDK bundles its own CLI binary which is version-matched to the SDK.
-        Using a system-installed CLI can cause version mismatches and auth issues.
-        Only override cli_path if explicitly configured.
+        We always use the bundled CLI to avoid version mismatches and auth issues.
         """
         options: dict[str, Any] = {}
 
-        # Only use cli_path if explicitly configured (config or env var)
-        # Otherwise let SDK use its bundled CLI binary
-        cli_path = self._config.get("cli_path") or os.environ.get("COPILOT_CLI_PATH")
-
-        if cli_path:
-            options["cli_path"] = cli_path
-            logger.debug(f"[CLIENT] Using explicit CLI path: {cli_path}")
-        else:
-            logger.debug("[CLIENT] Using SDK's bundled CLI")
-
         if self._config.get("log_level"):
             options["log_level"] = self._config["log_level"]
 
@@ -447,7 +435,7 @@ async def create_session(
         # Disable infinite sessions for ephemeral pattern
         session_config["infinite_sessions"] = {"enabled": False}
 
-        # Add tools for structured tool calling (Option A: capture-and-abort)
+        # Add tools for structured tool calling (capture-and-abort pattern)
         if tools:
             session_config["tools"] = tools
             logger.debug(f"[CLIENT] Registering {len(tools)} tool(s) with session")