monkeybot-cli 0.9.0__tar.gz → 0.10.0__tar.gz

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/PKG-INFO

@@ -1,11 +1,11 @@
 Metadata-Version: 2.5
 Name: monkeybot-cli
-Version: 0.9.0
+Version: 0.10.0
 Summary: CLI to create, configure, validate, and chat with monkeybot agents.
 Requires-Python: >=3.11
 Requires-Dist: httpx>=0.27.0
 Requires-Dist: monkeybot-browser-mcp<1,>=0.5.0
-Requires-Dist: monkeybot[cli]<4,>=3.8.0
+Requires-Dist: monkeybot[cli]<4,>=3.9.0
 Requires-Dist: python-dotenv>=1.0.0
 Requires-Dist: pyyaml>=6.0.2
 Requires-Dist: textual>=8.2.8

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/pyproject.toml

@@ -4,13 +4,13 @@ build-backend = "hatchling.build"
 
 [project]
 name = "monkeybot-cli"
-version = "0.9.0"
+version = "0.10.0"
 description = "CLI to create, configure, validate, and chat with monkeybot agents."
 requires-python = ">=3.11"
 dependencies = [
     # `cli` pulls realtime + typer so `monkeybot talk` works out of the box (text mode).
     # Audio mic/PTT needs the agent (or this env) to also install monkeybot[cli-realtime].
-    "monkeybot[cli]>=3.8.0,<4",
+    "monkeybot[cli]>=3.9.0,<4",
     "monkeybot-browser-mcp>=0.5.0,<1",
     "httpx>=0.27.0",
     "pyyaml>=6.0.2",

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/skills/monkeybot/references/config-sections.md

@@ -2,7 +2,7 @@
 
 Deep reference for every `monkeybot.yaml` section. Load this only when a user needs to customize beyond Tier 1. The canonical, fully-commented template lives at `cli/src/monkeybot_cli/scaffold_defaults/monkeybot.example.yaml` in the monkeybot repo (copied to `monkeybot_config/monkeybot.example.yaml` when you scaffold); this file adds the **"when would I change this?"** context the comments don't.
 
-**Precedence:** env vars and `.env` win over YAML for `ENV_MAP` keys. `model.*` (provider, name, temperature, max_tokens, thinking_budget, context_window, summarization_model, max_turns, cache_retention) is YAML-only — leftover `MODEL_*` env vars are ignored. The YAML→env mapping is `ENV_MAP` in `src/monkeybot/core/config/runtime_env.py`. If a YAML edit has no effect, check for a shadowing env var (except `model.*`).
+**Precedence:** env vars and `.env` win over YAML for `ENV_MAP` keys. `model.*` (provider, name, temperature, max_tokens, thinking_budget, reasoning_effort, context_window, summarization_model, max_turns, cache_retention) is YAML-only — leftover `MODEL_*` env vars are ignored. The YAML→env mapping is `ENV_MAP` in `src/monkeybot/core/config/runtime_env.py`. If a YAML edit has no effect, check for a shadowing env var (except `model.*`).
 
 ---
 
@@ -42,7 +42,8 @@ YAML-only for provider, name, sampling, context window, summarization model, max
 | `name` | `gemini-3-flash` | Pick a specific model id |
 | `temperature` | `0.7` | Lower for deterministic output, higher for creative |
 | `max_tokens` | `60000` | Cap per-response length |
-| `thinking_budget` | `-1` | Gemini: `-1` model default, `0` off, `N` token budget. Ollama reasoning models: `-1` server default, `0` off (`reasoning_effort: none`) |
+| `thinking_budget` | `-1` | Gemini: `-1` model default, `0` off, `N` token budget. Ollama reasoning models: `-1` server default, `0` off (`reasoning_effort: none`). Ignored when `reasoning_effort` is set on providers that take an effort level |
+| `reasoning_effort` | (unset) | Effort level passed through to the provider (`low`, `medium`, `high`, `xhigh`, `max`, or whatever that provider accepts). Omit for the provider default. YAML-only. Gemini only sends SDK levels (`minimal`, `low`, `medium`, `high`) and otherwise uses `thinking_budget`. A level the API rejects is omitted on later turns. Bedrock sends it for Claude ids only; Converse models ignore it. Curator and summarizer calls skip the configured effort; when one is set, OpenAI and Ollama send `none` instead |
 | `context_window` | `1000000` | Summarization trigger (tokens); also drives soft-spill / `read_file` char budgets. **Not** Ollama `num_ctx` |
 | `keep_alive` | `24h` (ollama-local only) | How long local Ollama keeps the model (and KV prefix cache) loaded. YAML only. Set `"0"` to omit. See `docs/ollama-local.md` |
 | `num_ctx` | (unset) | Optional pinned Ollama `num_ctx`. YAML only. Omit = server default. Do not copy `context_window` |

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/chat_session.py

@@ -101,9 +101,7 @@ def extract_elicitation_fields(payload: dict[str, Any]) -> tuple[str | None, dic
     else:
         message = message.strip()
     schema_raw = (
-        payload.get("requestedSchema")
-        or payload.get("requested_schema")
-        or payload.get("schema")
+        payload.get("requestedSchema") or payload.get("requested_schema") or payload.get("schema")
     )
     schema = dict(schema_raw) if isinstance(schema_raw, dict) else None
     return message, schema
@@ -662,9 +660,7 @@ class ChatSessionController:
 
         self._maybe_thinking_trace(evt, request_id)
 
-    def _on_context_usage_hint(
-        self, *, estimated_tokens: int, context_window_tokens: int
-    ) -> None:
+    def _on_context_usage_hint(self, *, estimated_tokens: int, context_window_tokens: int) -> None:
         self.usage.update_context_hint(
             estimated_prompt_tokens=estimated_tokens,
             context_window_tokens=context_window_tokens,
@@ -750,9 +746,7 @@ class ChatSessionController:
         if answer.cancelled:
             approved = False
         reason = (
-            "cancelled by user"
-            if answer.cancelled
-            else (None if approved else "denied by user")
+            "cancelled by user" if answer.cancelled else (None if approved else "denied by user")
         )
         ok = await self._post_hitl(
             f"{self.base}/sessions/{self.session_id}/tool-confirmations/{evt.tool_call_id}",

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/chat_tool_display.py

@@ -37,9 +37,7 @@ _SHELL_TAIL_LINES = 40
 
 _SHELL_TOOLS = frozenset({"run_command", "execute", "shell", "bash"})
 _READ_TOOLS = frozenset({"read_file", "read"})
-_EDIT_TOOLS = frozenset(
-    {"write_file", "write", "edit_file", "apply_patch", "str_replace"}
-)
+_EDIT_TOOLS = frozenset({"write_file", "write", "edit_file", "apply_patch", "str_replace"})
 _PATH_TOOLS = _READ_TOOLS | _EDIT_TOOLS
 _SEARCH_TOOLS = frozenset({"grep", "web_search"})
 _HINT_KEYS = frozenset(

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/chat_tui.py

@@ -93,6 +93,7 @@ def cycle_approval_mode(mode: str) -> str:
         idx = 0
     return _APPROVAL_MODES[(idx + 1) % len(_APPROVAL_MODES)]
 
+
 _SLASH_SPECS: tuple[tuple[str, str], ...] = (
     ("/help", "Show commands and key hints"),
     ("/new", "Start a fresh session"),
@@ -327,11 +328,7 @@ class ChatApp(App[int]):
         self.theme = resolve_theme_name(theme_choice)
         self.base = base
         self.agent_root = agent_root
-        self.config_path = (
-            config_path.expanduser().resolve()
-            if config_path is not None
-            else None
-        )
+        self.config_path = config_path.expanduser().resolve() if config_path is not None else None
         self.provider = provider
         self.model = model
         self.spawned_gateway = spawned_gateway
@@ -443,9 +440,7 @@ class ChatApp(App[int]):
             else:
                 parts.append("y approve · n deny · Ctrl-C cancel")
         elif self.show_hints:
-            parts.append(
-                "Enter send · Ctrl+J newline · ↑ history · / commands · Ctrl+U usage"
-            )
+            parts.append("Enter send · Ctrl+J newline · ↑ history · / commands · Ctrl+U usage")
         else:
             parts.append("F1 hints")
         return "  ·  ".join(parts)
@@ -495,8 +490,7 @@ class ChatApp(App[int]):
             self._session_cost_text = f"${usage.cost_usd:.4f}"
         if self.show_usage:
             self._usage_text = (
-                f"in={usage.input_tokens} out={usage.output_tokens} "
-                f"${usage.cost_usd:.4f}"
+                f"in={usage.input_tokens} out={usage.output_tokens} ${usage.cost_usd:.4f}"
             )
 
     def _hide_empty_hint(self) -> None:
@@ -556,6 +550,7 @@ class ChatApp(App[int]):
             if transcript.is_anchored and not transcript._anchor_released:
                 transcript.release_anchor()
         self._refresh_jump_button()
+
     def _refresh_jump_button(self) -> None:
         with contextlib.suppress(NoMatches):
             btn = self.query_one("#jump-bottom", Button)
@@ -627,7 +622,12 @@ class ChatApp(App[int]):
                 if isinstance(child, EarlierTurns):
                     earlier = child
                     continue
-                if child is self._assistant or child is self._thinking or child is self._thinking_trace or child is self._hitl_card:
+                if (
+                    child is self._assistant
+                    or child is self._thinking
+                    or child is self._thinking_trace
+                    or child is self._hitl_card
+                ):
                     continue
                 if isinstance(child, ToolCallBlock) and (
                     child in self._open_tools.values() or child in self._anon_tools
@@ -683,6 +683,7 @@ class ChatApp(App[int]):
     def _seal_thinking_trace(self) -> None:
         """Finish the current thinking block and start fresh on the next delta."""
         self._finish_thinking_trace(clear=True)
+
     def _mount_user(self, body: str) -> None:
         self._mount(UserTurn(body, show_timestamp=self.show_timestamps))
 
@@ -768,9 +769,7 @@ class ChatApp(App[int]):
             palette.display = False
             palette.clear_options()
             return
-        palette.set_options(
-            [Option(f"{cmd}  {desc}", id=cmd) for cmd, desc in matches]
-        )
+        palette.set_options([Option(f"{cmd}  {desc}", id=cmd) for cmd, desc in matches])
         palette.highlighted = 0
         palette.display = True
 
@@ -945,9 +944,7 @@ class ChatApp(App[int]):
         if state == "reconnecting":
             attempt = p.get("attempt")
             self._conn_text = (
-                f"[yellow]reconnecting…[/] ({attempt})"
-                if attempt
-                else "[yellow]reconnecting…[/]"
+                f"[yellow]reconnecting…[/] ({attempt})" if attempt else "[yellow]reconnecting…[/]"
             )
         elif state == "connected":
             self._conn_text = ""
@@ -1508,7 +1505,9 @@ class ChatApp(App[int]):
         self.model = name
         self._refresh_topbar()
         await self._reset_transcript_ui()
-        self._mount_system(f"Model set to {provider}/{name} — new session started (context cleared)")
+        self._mount_system(
+            f"Model set to {provider}/{name} — new session started (context cleared)"
+        )
         self._refresh_status()
         self.query_one("#prompt", Composer).focus()
 
@@ -1566,6 +1565,7 @@ class ChatApp(App[int]):
         except Exception as exc:  # noqa: BLE001 - surface any YAML/parse error to the user
             self._mount_system(f"Could not read config: {exc}", error=True)
             return
+
         def _section(key: str) -> dict[str, object]:
             value = doc.get(key)
             return value if isinstance(value, dict) else {}
@@ -1755,7 +1755,6 @@ class ChatApp(App[int]):
     def action_toggle_usage(self) -> None:
         self._cmd_usage()
 
-
     def action_toggle_ptt(self) -> None:
         """Space toggles in-TUI PTT when composer is empty (realtime / SSH)."""
         with contextlib.suppress(NoMatches):
@@ -1786,10 +1785,7 @@ class ChatApp(App[int]):
             self._controller.abort_turn()
             return
         now = time.monotonic()
-        if (
-            self._last_esc is not None
-            and now - self._last_esc <= _ESC_RECALL_WINDOW
-        ):
+        if self._last_esc is not None and now - self._last_esc <= _ESC_RECALL_WINDOW:
             self._last_esc = None
             composer.recall_last()
             return
@@ -1815,11 +1811,7 @@ class ChatApp(App[int]):
         self._goodbye_and_exit()
 
     def _goodbye_and_exit(self) -> None:
-        msg = (
-            "Goodbye — shutting down gateway…"
-            if self.spawned_gateway
-            else "Goodbye."
-        )
+        msg = "Goodbye — shutting down gateway…" if self.spawned_gateway else "Goodbye."
         self._mount_system(msg)
         if self._session_id:
             self._mount_system("To continue this conversation, run: monkeybot chat --continue")

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/commands/chat.py

@@ -64,6 +64,7 @@ _UNDERLINE = "\x1b[4m"
 _SPINNER_FRAMES = "⠋⠙⠹⠸⠼⠴⠦⠧⠇⠏"
 _GROUNDING_SOURCES_MAX = 5
 
+
 def _hyperlink(url: str, label: str) -> str:
     """OSC 8 terminal hyperlink; falls back to a plain URL on non-TTY streams."""
     if not url:
@@ -278,9 +279,7 @@ def _resolve_base_url(args: argparse.Namespace, config_path: Path | None) -> str
     return f"http://127.0.0.1:{port}"
 
 
-def _model_banner_fields(
-    args: argparse.Namespace, config_path: Path | None
-) -> tuple[str, str]:
+def _model_banner_fields(args: argparse.Namespace, config_path: Path | None) -> tuple[str, str]:
     provider = (args.model_provider or "").strip()
     model = (args.model_name or "").strip()
     if config_path is not None and (not provider or not model):

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/commands/doctor.py

@@ -10,9 +10,12 @@ import tomllib
 from pathlib import Path
 
 import httpx
+from monkeybot.core.config.settings import normalize_model_provider
 from monkeybot.core.layout import AgentLayout, bootstrap_agent_layout
 from monkeybot.core.memory.config import memory_enabled_from_config
 from monkeybot.core.tools.sandbox_executor import SandboxConfig
+from monkeybot.providers._bedrock_converse import uses_anthropic_bedrock
+from monkeybot.providers.sampling import normalize_reasoning_effort
 
 from monkeybot_cli.compat import COMPATIBLE_CORE_RANGE
 from monkeybot_cli.config_resolve import (
@@ -228,6 +231,58 @@ def _strict_positive_int(raw: object) -> tuple[int | None, bool]:
     return raw, False
 
 
+def _add_reasoning_effort_check(
+    report: CommandReport,
+    *,
+    reasoning_effort: object,
+    provider: str = "",
+    model_name: str = "",
+) -> None:
+    """Note a YAML effort value.
+
+    Severity stays ``warning``: the doctor schema has no ``info`` level, and a
+    passing check does not increment the warning summary. A non-Claude Bedrock
+    model fails this check because the Converse path does not send the field.
+    """
+    text = normalize_reasoning_effort(reasoning_effort) or ""
+    converse = (
+        normalize_model_provider(provider) in {"aws_bedrock", "bedrock"}
+        and bool(model_name)
+        and not uses_anthropic_bedrock(model_name)
+    )
+    if text == "none":
+        message = "reasoning_effort turns thinking off where the provider supports it."
+        passed = True
+    elif text and converse:
+        message = (
+            f"reasoning_effort {text!r} is not sent for this Bedrock model. "
+            "Only Claude ids receive it; Converse models ignore it."
+        )
+        passed = False
+    elif text:
+        message = (
+            f"reasoning_effort {text!r} is sent to the provider. "
+            "It wins over thinking_budget where the provider accepts an effort level. "
+            "Gemini only sends levels its SDK knows; other values use thinking_budget. "
+            "A level the API rejects is omitted on later turns."
+        )
+        passed = True
+    else:
+        message = "reasoning_effort unset"
+        passed = True
+    check(
+        report,
+        id="model.reasoning_effort",
+        category="provider",
+        severity="warning",
+        passed=passed,
+        skip=not text,
+        message=message,
+        field="model.reasoning_effort",
+        value=text or None,
+    )
+
+
 def _add_ollama_local_checks(
     report: CommandReport,
     *,
@@ -464,11 +519,19 @@ def run_doctor(args: argparse.Namespace) -> int:
             skip=True,
         )
 
+    effort_raw = model.get("reasoning_effort") if isinstance(model, dict) else None
+    model_name = str(model.get("name", "")).strip() if isinstance(model, dict) else ""
+    _add_reasoning_effort_check(
+        report,
+        reasoning_effort=effort_raw,
+        provider=provider,
+        model_name=model_name,
+    )
+
     thinking_raw = model.get("thinking_budget") if isinstance(model, dict) else None
     if thinking_raw is None:
         thinking_raw = "-1"
     num_ctx_raw = model.get("num_ctx") if isinstance(model, dict) else None
-    model_name = str(model.get("name", "")).strip() if isinstance(model, dict) else ""
     _add_ollama_local_checks(
         report,
         provider=provider,

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/commands/loop.py

@@ -9,7 +9,12 @@ from pathlib import Path
 
 import httpx
 
-from monkeybot_cli.config_resolve import load_agent_dotenv, load_config_doc, resolve_agent_root, resolve_config
+from monkeybot_cli.config_resolve import (
+    load_agent_dotenv,
+    load_config_doc,
+    resolve_agent_root,
+    resolve_config,
+)
 
 DEFAULT_GATEWAY_PORT = 8000
 

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/commands/new.py

@@ -144,9 +144,7 @@ def run_new_command(args: argparse.Namespace) -> int:
                 default_key=_DEFAULT_PROVIDER,
             )
         if model is None:
-            model = (
-                _prompt_line(f"Model name [{_DEFAULT_MODEL}]: ").strip() or _DEFAULT_MODEL
-            )
+            model = _prompt_line(f"Model name [{_DEFAULT_MODEL}]: ").strip() or _DEFAULT_MODEL
         # Only prompt for optional deps when --with was not already given.
         if not extras:
             more_providers = additional_provider_extra_choices(provider)

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/commands/validate.py

@@ -120,10 +120,19 @@ def run_validate(args: argparse.Namespace) -> int:
             category="config",
             severity="warning",
             passed=not missing,
-            message=f"Missing includes: {', '.join(missing)}" if missing else "All includes resolved",
+            message=f"Missing includes: {', '.join(missing)}"
+            if missing
+            else "All includes resolved",
         )
     else:
-        check(report, id="config.includes.resolve", category="config", severity="warning", passed=True, skip=True)
+        check(
+            report,
+            id="config.includes.resolve",
+            category="config",
+            severity="warning",
+            passed=True,
+            skip=True,
+        )
 
     model = merged.get("model") if isinstance(merged.get("model"), dict) else {}
     provider = str(model.get("provider", "")).strip().lower() if isinstance(model, dict) else ""
@@ -134,7 +143,9 @@ def run_validate(args: argparse.Namespace) -> int:
         category="config",
         severity="error",
         passed=not provider or provider in SUPPORTED_YAML_MODEL_PROVIDERS,
-        message=f"model.provider '{provider}' is not supported" if provider and provider not in SUPPORTED_YAML_MODEL_PROVIDERS else "",
+        message=f"model.provider '{provider}' is not supported"
+        if provider and provider not in SUPPORTED_YAML_MODEL_PROVIDERS
+        else "",
         field="model.provider",
         value=provider,
         expected=sorted(SUPPORTED_YAML_MODEL_PROVIDERS),
@@ -172,12 +183,23 @@ def run_validate(args: argparse.Namespace) -> int:
     )
 
     flat = _flatten_to_env(merged)
-    flat.update({k: v for k, v in os.environ.items() if k.startswith(("GCP_", "GOOGLE_", "ANTHROPIC_VERTEX"))})
+    flat.update(
+        {
+            k: v
+            for k, v in os.environ.items()
+            if k.startswith(("GCP_", "GOOGLE_", "ANTHROPIC_VERTEX"))
+        }
+    )
     norm_provider = normalize_model_provider(provider or "gemini")
     needs_gcp = norm_provider == "vertex_anthropic"
     gcp_present = any(
         os.environ.get(k, "").strip() or flat.get(k, "").strip()
-        for k in ("GCP_PROJECT_ID", "VERTEX_AI_PROJECT_ID", "ANTHROPIC_VERTEX_PROJECT_ID", "GOOGLE_CLOUD_PROJECT")
+        for k in (
+            "GCP_PROJECT_ID",
+            "VERTEX_AI_PROJECT_ID",
+            "ANTHROPIC_VERTEX_PROJECT_ID",
+            "GOOGLE_CLOUD_PROJECT",
+        )
     )
     if needs_gcp:
         check(
@@ -190,11 +212,20 @@ def run_validate(args: argparse.Namespace) -> int:
             remediation="Set gcp.project_id in monkeybot.yaml or GCP_PROJECT_ID in .env",
         )
     else:
-        check(report, id="gcp.project.required", category="config", severity="error", passed=True, skip=True)
+        check(
+            report,
+            id="gcp.project.required",
+            category="config",
+            severity="error",
+            passed=True,
+            skip=True,
+        )
 
     try:
         validate_monkeybot_yaml_doc(merged, env=dict(os.environ))
-        validate_provider_env({**flat, "MODEL_PROVIDER": provider or flat.get("MODEL_PROVIDER", "gemini")})
+        validate_provider_env(
+            {**flat, "MODEL_PROVIDER": provider or flat.get("MODEL_PROVIDER", "gemini")}
+        )
     except ConfigError as exc:
         check(
             report,
@@ -329,7 +360,9 @@ def run_validate(args: argparse.Namespace) -> int:
                         category="mcp",
                         severity="warning",
                         passed=not missing_refs,
-                        message=f"Unresolved MCP env refs: {', '.join(missing_refs)}" if missing_refs else "All MCP env refs set",
+                        message=f"Unresolved MCP env refs: {', '.join(missing_refs)}"
+                        if missing_refs
+                        else "All MCP env refs set",
                     )
                     if args.check_mcp:
                         for name, srv in servers.items():

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/compat.py

@@ -14,14 +14,10 @@ def _bounds_from_range(range_: str) -> tuple[str, str]:
     """Parse inclusive lower / exclusive upper versions from ``COMPATIBLE_CORE_RANGE``."""
     match = re.fullmatch(r">\s*=\s*([^,]+)\s*,\s*<\s*(.+)", range_.strip())
     if match is None:
-        raise ValueError(
-            f"COMPATIBLE_CORE_RANGE must be '>=LOWER,<UPPER', got {range_!r}"
-        )
+        raise ValueError(f"COMPATIBLE_CORE_RANGE must be '>=LOWER,<UPPER', got {range_!r}")
     lower, upper = match.group(1).strip(), match.group(2).strip()
     if not lower or not upper:
-        raise ValueError(
-            f"COMPATIBLE_CORE_RANGE must be '>=LOWER,<UPPER', got {range_!r}"
-        )
+        raise ValueError(f"COMPATIBLE_CORE_RANGE must be '>=LOWER,<UPPER', got {range_!r}")
     return lower, upper
 
 

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/extras_catalog.py

@@ -28,6 +28,7 @@ PROVIDER_CHOICES: tuple[ExtraChoice, ...] = (
     ExtraChoice("ollama-local", "Ollama (local)"),
     ExtraChoice("nvidia", "NVIDIA NIM"),
     ExtraChoice("openrouter", "OpenRouter"),
+    ExtraChoice("fireworks", "Fireworks AI"),
 )
 
 # Non-provider agent features (root ``[project.optional-dependencies]`` names).

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/gateway_health.py

@@ -20,9 +20,7 @@ def port_free(port: int, host: str = "127.0.0.1") -> bool:
             return False
 
 
-def wait_for_health(
-    base: str, proc: subprocess.Popen[str] | None, timeout_s: float = 30.0
-) -> bool:
+def wait_for_health(base: str, proc: subprocess.Popen[str] | None, timeout_s: float = 30.0) -> bool:
     """Poll ``GET {base}/health`` until 200 from *our* child, process dies, or timeout.
 
     On success we re-check ``proc.poll()`` so a pre-existing listener on the same

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/opensandbox_lifecycle.py

@@ -261,7 +261,9 @@ def ensure_opensandbox_for_agent(
         )
         return False
 
-    container = os.environ.get("SANDBOX_CONTAINER", _DEFAULT_CONTAINER).strip() or _DEFAULT_CONTAINER
+    container = (
+        os.environ.get("SANDBOX_CONTAINER", _DEFAULT_CONTAINER).strip() or _DEFAULT_CONTAINER
+    )
     image = _server_image()
     if health_wait_secs is None:
         health_wait_secs = float(
@@ -312,8 +314,7 @@ def ensure_opensandbox_for_agent(
 
     if _container_exists(container) and not _container_running(container):
         print(
-            "monkeybot: OpenSandbox container exited before becoming healthy "
-            f"(image={image}).",
+            f"monkeybot: OpenSandbox container exited before becoming healthy (image={image}).",
             flush=True,
         )
     else:

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/output.py

@@ -73,7 +73,9 @@ class CommandReport:
             if c.remediation and c.status == "fail":
                 print(f"      → {c.remediation}")
         s = self.summary
-        print(f"  summary: {s['passed']} passed, {s['warnings']} warnings, {s['failed']} failed, {s['skipped']} skipped")
+        print(
+            f"  summary: {s['passed']} passed, {s['warnings']} warnings, {s['failed']} failed, {s['skipped']} skipped"
+        )
 
     def emit(self, *, as_json: bool) -> int:
         self.compute_ok()

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/providers.py

@@ -50,6 +50,7 @@ PROVIDER_SPECS: dict[str, ProviderSpec] = {
     ),
     "nvidia": ProviderSpec(("nvidia",), "nvidia", ("NVIDIA_API_KEY",)),
     "openrouter": ProviderSpec(("openrouter",), "openrouter", ("OPENROUTER_API_KEY",)),
+    "fireworks": ProviderSpec(("fireworks",), "fireworks", ("FIREWORKS_API_KEY",)),
     "ollama": ProviderSpec(
         # credential_env_vars is empty: credentials_optional=True short-circuits
         # credentials_present() before these are ever read, and OLLAMA_BASE_URL
@@ -96,6 +97,7 @@ def extra_module(extra: str) -> str:
         "ollama": "openai",
         "nvidia": "openai",
         "openrouter": "openai",
+        "fireworks": "openai",
         "postgres": "asyncpg",
         "firestore": "google.cloud.firestore",
         "gcs": "google.cloud.storage",

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/realtime/push_to_talk.py

@@ -57,8 +57,7 @@ class PushToTalkGate:
         keys = aliases.get(key_name)
         if keys is None:
             raise PushToTalkError(
-                f"Unsupported push-to-talk key '{key_name}'. "
-                "Use one of: cmd, alt, ctrl, space"
+                f"Unsupported push-to-talk key '{key_name}'. Use one of: cmd, alt, ctrl, space"
             )
         return keys
 

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/realtime/session_controller.py

@@ -171,9 +171,7 @@ class RealtimeSessionController:
                 self.push_to_talk = None
         try:
             self._ws = await websockets.connect(url)
-            await self._ws.send(
-                json.dumps({"kind": "connect", "session_id": self.session_id})
-            )
+            await self._ws.send(json.dumps({"kind": "connect", "session_id": self.session_id}))
         except websockets.exceptions.InvalidStatus as exc:
             status = getattr(exc, "status_code", getattr(exc, "status", "unknown"))
             raise RuntimeError(f"Failed to connect to {url}: HTTP {status}") from exc

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/scaffold_defaults/env.example

@@ -28,6 +28,9 @@ DB_URL=sqlite:///data/monkeybot.db
 # Hugging Face
 # HF_TOKEN=
 
+# Fireworks AI
+# FIREWORKS_API_KEY=
+
 # Ollama Cloud (model.provider: ollama-cloud) — key from ollama.com/settings/keys
 # OLLAMA_API_KEY=
 

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/src/monkeybot_cli/scaffold_defaults/monkeybot.example.yaml

@@ -4,7 +4,8 @@
 #
 # Precedence: environment variables (and optional root `.env`) win over values here
 # except ``model.*`` (provider, name, temperature, max_tokens, thinking_budget,
-# context_window, summarization_model, max_turns, cache_retention) which are YAML-only.
+# reasoning_effort, context_window, summarization_model, max_turns, cache_retention)
+# which are YAML-only.
 #
 # Secrets are not committed here. Create a `.env` in the repo root when you need keys or
 # machine-local paths, for example:
@@ -14,6 +15,7 @@
 #   ANTHROPIC_VERTEX_PROJECT_ID, ANTHROPIC_VERTEX_REGION,
 #   OLLAMA_BASE_URL (default http://localhost:11434, no API key needed),
 #   NVIDIA_API_KEY (free key from https://build.nvidia.com),
+#   FIREWORKS_API_KEY (https://fireworks.ai/account/api-keys),
 #   TAVILY_API_KEY, FIRECRAWL_API_KEY, SANDBOX_API_KEY,
 #   MONKEYBOT_TOOL_DENIED_PATTERNS, MONKEYBOT_CORS_ALLOW_ORIGINS, MONKEYBOT_CONFIG, …
 #
@@ -89,7 +91,7 @@ paths:
   workspace_root: ./workspace
 
 model:
-  # gemini | openai | anthropic | vertex-claude | huggingface | ollama-cloud | ollama-local | ollama | nvidia | openrouter | aws_bedrock | fake
+  # gemini | openai | anthropic | vertex-claude | huggingface | ollama-cloud | ollama-local | ollama | nvidia | openrouter | fireworks | aws_bedrock | fake
   provider: gemini
   name: gemini-3.1-flash-live-preview
   temperature: 0.7
@@ -97,6 +99,13 @@ model:
   # Gemini: -1 = model default, 0 = off, N = token budget
   # Ollama (Gemma 4, Qwen3, …): -1 = server default, 0 = off (reasoning_effort: none)
   thinking_budget: -1
+  # Effort level for models that take one (low, medium, high, xhigh, max, …).
+  # Omit to use the provider default. Wins over thinking_budget when set.
+  # off / none / disabled turn thinking off (a bare YAML `off` is read as false; also off).
+  # Gemini only sends levels its SDK knows (minimal, low, medium, high); anything
+  # else falls back to thinking_budget. A value the API rejects is omitted on
+  # later turns. Bedrock sends it for Claude ids only.
+  # reasoning_effort: high
   # Summarisation trigger (tokens). Also drives soft-spill / read_file char budgets.
   context_window: 1000000
   max_turns: 1000

{monkeybot_cli-0.9.0 → monkeybot_cli-0.10.0}/tests/test_chat_continue.py

@@ -133,9 +133,7 @@ def test_run_chat_continue_resolves_and_sets_session(
         "monkeybot_cli.commands.chat._resolve_continue_session_id",
         lambda base: "resolved-session-id",
     )
-    args = build_parser().parse_args(
-        ["chat", "--url", "http://127.0.0.1:65001", "--continue"]
-    )
+    args = build_parser().parse_args(["chat", "--url", "http://127.0.0.1:65001", "--continue"])
     run_chat(args)
     assert args.session == "resolved-session-id"
     assert "Continuing session resolved" in capsys.readouterr().out