zwarm 3.9.0__tar.gz → 3.10.2__tar.gz

{zwarm-3.9.0 → zwarm-3.10.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: zwarm
-Version: 3.9.0
+Version: 3.10.2
 Summary: Multi-Agent CLI Orchestration Research Platform
 Requires-Python: <3.14,>=3.13
 Requires-Dist: prompt-toolkit>=3.0.52
@@ -8,6 +8,7 @@ Requires-Dist: python-dotenv>=1.0.0
 Requires-Dist: pyyaml>=6.0
 Requires-Dist: rich>=13.0.0
 Requires-Dist: typer>=0.9.0
+Requires-Dist: wandb>=0.24.0
 Requires-Dist: wbal>=0.5.8
 Description-Content-Type: text/markdown
 

{zwarm-3.9.0 → zwarm-3.10.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "zwarm"
-version = "3.9.0"
+version = "3.10.2"
 description = "Multi-Agent CLI Orchestration Research Platform"
 readme = "README.md"
 requires-python = ">=3.13,<3.14"
@@ -11,6 +11,7 @@ dependencies = [
     "pyyaml>=6.0",
     "wbal>=0.5.8",
     "prompt-toolkit>=3.0.52",
+    "wandb>=0.24.0",
 ]
 
 [project.scripts]

{zwarm-3.9.0 → zwarm-3.10.2}/src/zwarm/cli/interactive.py

@@ -157,7 +157,7 @@ def cmd_help():
     table.add_column("Description")
 
     table.add_row("[bold]Session Lifecycle[/]", "")
-    table.add_row('spawn "task" [--model M] [--adapter A]', "Start new session")
+    table.add_row('spawn "task" [--model M]', "Start new session")
     table.add_row('c ID "message"', "Continue conversation")
     table.add_row("kill ID | all", "Stop session(s)")
     table.add_row("rm ID | all", "Delete session(s)")
@@ -832,7 +832,7 @@ def cmd_rm(manager, target: str):
 
 def run_interactive(
     working_dir: Path,
-    model: str = "gpt-5.1-codex-mini",
+    model: str = "gpt-5.2-codex",
 ):
     """
     Run the interactive REPL.
@@ -1020,7 +1020,8 @@ def run_interactive(
 
             elif cmd == "spawn":
                 if not args:
-                    console.print("  [red]Usage:[/] spawn \"task\" [--model M] [--adapter A]")
+                    console.print("  [red]Usage:[/] spawn \"task\" [--model M]")
+                    console.print("  [dim]Models: 5.2, 5.2-think, opus, sonnet (adapter auto-detected)[/]")
                 else:
                     # Parse spawn args
                     task_parts = []

{zwarm-3.9.0 → zwarm-3.10.2}/src/zwarm/cli/main.py

@@ -441,7 +441,7 @@ def exec(
         console.print(f"  Model: {model}")
 
     manager = CodexSessionManager(working_dir / ".zwarm")
-    effective_model = model or "gpt-5.1-codex-mini"
+    effective_model = model or "gpt-5.2-codex"
 
     session = manager.start_session(
         task=task,
@@ -813,7 +813,7 @@ def init(
     project_description = ""
     project_context = ""
     # Codex settings
-    codex_model = "gpt-5.1-codex-mini"
+    codex_model = "gpt-5.2-codex"
     codex_reasoning = "high"
 
     if not non_interactive:
@@ -838,9 +838,9 @@ def init(
         console.print("  [dim]These control the underlying Codex CLI that runs executor sessions[/]\n")
 
         console.print("  Available models:")
-        console.print("    [cyan]1[/] gpt-5.1-codex-mini  [dim]- Fast, cheap, good for most tasks (Recommended)[/]")
-        console.print("    [cyan]2[/] gpt-5.1-codex       [dim]- Balanced speed and capability[/]")
-        console.print("    [cyan]3[/] gpt-5.1-codex-max   [dim]- Most capable, 400k context, expensive[/]")
+        console.print("    [cyan]1[/] gpt-5.2-codex   [dim]- GPT-5.2 Codex, balanced (Recommended)[/]")
+        console.print("    [cyan]2[/] gpt-5.2         [dim]- GPT-5.2 with extended reasoning[/]")
+        console.print("    [cyan]3[/] gpt-5.1-codex   [dim]- GPT-5.1 Codex (legacy)[/]")
 
         model_choice = typer.prompt(
             "  Select model (1-3)",
@@ -848,9 +848,9 @@ def init(
             type=str,
         )
         model_map = {
-            "1": "gpt-5.1-codex-mini",
-            "2": "gpt-5.1-codex",
-            "3": "gpt-5.1-codex-max",
+            "1": "gpt-5.2-codex",
+            "2": "gpt-5.2",
+            "3": "gpt-5.1-codex",
         }
         codex_model = model_map.get(model_choice, model_choice)
         if model_choice not in model_map:
@@ -1041,7 +1041,7 @@ def _generate_config_toml(
         "# ============================================================================",
         "[executor]",
         f'adapter = "{adapter}"                  # codex_mcp | codex_exec | claude_code',
-        '# model = "gpt-5.1-codex-mini"         # Model for delegated sessions (uses codex.toml default if not set)',
+        '# model = "gpt-5.2-codex"              # Model for delegated sessions (uses codex.toml default if not set)',
         'sandbox = "workspace-write"            # read-only | workspace-write | danger-full-access',
         "timeout = 3600                         # Session timeout in seconds",
         'reasoning_effort = "high"              # low | medium | high',
@@ -1082,7 +1082,7 @@ def _generate_config_toml(
 
 
 def _generate_codex_toml(
-    model: str = "gpt-5.1-codex-mini",
+    model: str = "gpt-5.2-codex",
     reasoning_effort: str = "high",
 ) -> str:
     """
@@ -1461,7 +1461,7 @@ def interactive(
     """
     from zwarm.cli.interactive import run_interactive
 
-    default_model = model or "gpt-5.1-codex-mini"
+    default_model = model or "gpt-5.2-codex"
     run_interactive(working_dir=default_dir.absolute(), model=default_model)
 
 
@@ -1655,7 +1655,7 @@ app.add_typer(session_app, name="session")
 def session_start(
     task: Annotated[str, typer.Argument(help="Task description")],
     working_dir: Annotated[Path, typer.Option("--dir", "-d", help="Working directory")] = Path("."),
-    model: Annotated[str, typer.Option("--model", "-m", help="Model to use")] = "gpt-5.1-codex-mini",
+    model: Annotated[str, typer.Option("--model", "-m", help="Model to use")] = "gpt-5.2-codex",
 ):
     """
     Start a new Codex session in the background.

{zwarm-3.9.0 → zwarm-3.10.2}/src/zwarm/cli/pilot.py

@@ -84,9 +84,10 @@ class ChoogingSpinner:
 # These are for the ORCHESTRATOR LLM, not the executors
 MODEL_CONTEXT_WINDOWS = {
     # OpenAI models
+    "gpt-5.2-codex": 200_000,
+    "gpt-5.2": 200_000,
     "gpt-5.1-codex": 200_000,
     "gpt-5.1-codex-mini": 200_000,
-    "gpt-5.1-codex-max": 400_000,
     "gpt-5": 200_000,
     "gpt-5-mini": 200_000,
     "o3": 200_000,

{zwarm-3.9.0 → zwarm-3.10.2}/src/zwarm/core/registry.py

@@ -61,34 +61,43 @@ MODELS: list[ModelInfo] = [
     # -------------------------------------------------------------------------
     # OpenAI Codex Models (via `codex` CLI)
     # -------------------------------------------------------------------------
+    ModelInfo(
+        canonical="gpt-5.2-codex",
+        adapter="codex",
+        aliases=["5.2", "5.2-codex"],
+        input_per_million=1.50,
+        output_per_million=12.00,
+        cached_input_per_million=0.15,
+        description="GPT-5.2 Codex - balanced speed/capability",
+        is_default=True,
+    ),
+    ModelInfo(
+        canonical="gpt-5.2",
+        adapter="codex",
+        aliases=["5.2-think", "gpt5.2"],
+        input_per_million=2.00,
+        output_per_million=16.00,
+        cached_input_per_million=0.20,
+        description="GPT-5.2 with extended reasoning (xhigh)",
+    ),
     ModelInfo(
         canonical="gpt-5.1-codex-mini",
         adapter="codex",
-        aliases=["codex-mini", "mini"],
+        aliases=["codex-mini", "mini", "5.1-mini"],
         input_per_million=0.25,
         output_per_million=2.00,
         cached_input_per_million=0.025,
         description="Fast, cost-effective coding model",
-        is_default=True,
     ),
     ModelInfo(
         canonical="gpt-5.1-codex",
         adapter="codex",
-        aliases=["codex", "codex-full"],
+        aliases=["codex", "codex-full", "5.1"],
         input_per_million=1.25,
         output_per_million=10.00,
         cached_input_per_million=0.125,
         description="Full Codex model with extended reasoning",
     ),
-    ModelInfo(
-        canonical="gpt-5.1-codex-max",
-        adapter="codex",
-        aliases=["codex-max", "max"],
-        input_per_million=1.25,
-        output_per_million=10.00,
-        cached_input_per_million=0.125,
-        description="Maximum context Codex model",
-    ),
     # -------------------------------------------------------------------------
     # Anthropic Claude Models (via `claude` CLI)
     # -------------------------------------------------------------------------
@@ -109,14 +118,6 @@ MODELS: list[ModelInfo] = [
         output_per_million=75.00,
         description="Most capable Claude model",
     ),
-    ModelInfo(
-        canonical="haiku",
-        adapter="claude",
-        aliases=["claude-haiku", "claude-4-haiku"],
-        input_per_million=0.25,
-        output_per_million=1.25,
-        description="Fast, lightweight Claude model",
-    ),
 ]
 
 

{zwarm-3.9.0 → zwarm-3.10.2}/src/zwarm/prompts/pilot.py

@@ -44,10 +44,13 @@ You command executor agents - capable coding agents that do specific tasks. Thin
 
 # Your Tools
 
-**delegate(task, adapter="codex", model=None, working_dir=None)** - Dispatch a crew member. Returns immediately with session_id.
-  - `adapter`: "codex" (fast, great for code) or "claude" (powerful reasoning)
-  - `model`: Override model (default: gpt-5.1-codex-mini for codex, sonnet for claude)
-  - Use codex for most tasks - it's fast. Use claude for complex reasoning.
+**delegate(task, model=None, working_dir=None)** - Dispatch a crew member. Returns immediately with session_id.
+  - `model`: Just use the model name - adapter is auto-detected:
+    - `"5.2"` or `"gpt-5.2-codex"` - GPT-5.2 Codex (default, fast, great for code)
+    - `"5.2-think"` - GPT-5.2 with extended reasoning
+    - `"opus"` - Claude Opus (most capable, complex reasoning)
+    - `"sonnet"` - Claude Sonnet (balanced)
+  - Use codex models for most tasks - they're fast. Use opus for complex reasoning.
 
 **converse(session_id, message)** - Send follow-up to a crew member. Returns immediately.
 
@@ -55,11 +58,16 @@ You command executor agents - capable coding agents that do specific tasks. Thin
 
 **check_session(session_id)** - Get FULL result. Complete response, tokens, runtime.
 
-**get_trajectory(session_id, full=False)** - See what steps the agent took (for debugging).
+**get_trajectory(session_id, full=False)** - See what steps the agent took.
+  - `full=True`: Show complete untruncated content for all steps (debugging)
+  - `full=False`: Concise summaries (default)
 
-**list_sessions()** - See all crew. `needs_attention=True` means ready for review.
+**list_sessions(status=None)** - See all crew. `needs_attention=True` means ready for review.
+  - `status`: Filter by "running", "completed", "failed", or None for all
 
-**end_session(session_id)** - Dismiss a crew member.
+**end_session(session_id, reason=None, delete=False)** - Dismiss a crew member.
+  - `reason`: Optional note about why
+  - `delete=True`: Permanently remove from list (otherwise just kills if running)
 
 **sleep(seconds)** - Wait before checking. Give crew time to work (15-60s typical).
 
@@ -68,7 +76,7 @@ You command executor agents - capable coding agents that do specific tasks. Thin
 # Workflow
 
 ```
-1. delegate(task) → session_id
+1. delegate(task, model="5.2") → session_id   # or model="opus" for complex tasks
 2. sleep(30)
 3. peek_session(id) → done?
 4. If running, goto 2

{zwarm-3.9.0 → zwarm-3.10.2}/src/zwarm/tools/delegation.py

@@ -163,17 +163,19 @@ def delegate(
     task: str,
     model: str | None = None,
     working_dir: str | None = None,
-    adapter: str = "codex",
+    adapter: str | None = None,
 ) -> dict[str, Any]:
     """
     Delegate work to an executor agent. Returns immediately - sessions run async.
 
-    Supports multiple adapters:
-    - codex: OpenAI's Codex CLI (default, fast, good for code tasks)
-    - claude: Claude Code CLI (powerful, good for complex reasoning)
+    Model shortcuts (adapter auto-detected):
+    - "5.2" or "gpt-5.2-codex": GPT-5.2 Codex (default, balanced)
+    - "5.2-think" or "gpt-5.2": GPT-5.2 with extended reasoning
+    - "sonnet": Claude Sonnet (balanced)
+    - "opus": Claude Opus (most capable)
 
     WORKFLOW:
-        1. delegate(task="...") -> session_id
+        1. delegate(task="...", model="5.2") -> session_id
         2. sleep(30)
         3. peek_session(session_id) -> {is_running: true/false}
         4. If is_running, goto 2
@@ -181,22 +183,35 @@ def delegate(
 
     Args:
         task: Clear description of what to do. Be specific about requirements.
-        model: Model override (codex: gpt-5.1-codex-mini, claude: sonnet).
+        model: Model name or alias. Adapter is auto-detected from model.
         working_dir: Directory for executor to work in (default: orchestrator's dir).
-        adapter: Which executor to use - "codex" (default) or "claude".
+        adapter: Override adapter (usually auto-detected from model).
 
     Returns:
-        {session_id, status: "running", adapter}
+        {session_id, status: "running", adapter, model}
 
     Example:
-        delegate(task="Add a logout button to the navbar")
-        delegate(task="Refactor auth to OAuth2", adapter="claude")
+        delegate(task="Add a logout button", model="5.2")
+        delegate(task="Complex refactor", model="opus")
     """
+    from zwarm.core.registry import get_adapter_for_model, resolve_model
+
+    # Auto-detect adapter from model if not specified
+    effective_adapter = adapter
+    if model and not adapter:
+        detected = get_adapter_for_model(model)
+        if detected:
+            effective_adapter = detected
+
+    # Fall back to codex if still no adapter
+    if not effective_adapter:
+        effective_adapter = "codex"
+
     # Validate adapter
-    if adapter not in ADAPTERS:
+    if effective_adapter not in ADAPTERS:
         return {
             "success": False,
-            "error": f"Unknown adapter: {adapter}. Available: {ADAPTERS}",
+            "error": f"Unknown adapter: {effective_adapter}. Available: {ADAPTERS}",
             "hint": f"Use one of: {ADAPTERS}",
         }
 
@@ -215,11 +230,12 @@ def delegate(
         }
 
     # Get the session manager for this adapter
-    manager = _get_adapter_manager(self, adapter)
+    manager = _get_adapter_manager(self, effective_adapter)
 
-    # Determine model (defaults vary by adapter)
+    # Determine model - resolve aliases to canonical names
     if model:
-        effective_model = model
+        model_info = resolve_model(model)
+        effective_model = model_info.canonical if model_info else model
     elif self.config.executor.model:
         effective_model = self.config.executor.model
     else:
@@ -245,7 +261,7 @@ def delegate(
         "session_id": session.id,
         "status": "running",
         "task": _truncate(task, 100),
-        "adapter": adapter,
+        "adapter": effective_adapter,
         "model": effective_model,
         "hint": "Use sleep() then check_session(session_id) to monitor progress",
     }