kolega-code 0.40.0__tar.gz → 0.41.1__tar.gz

{kolega_code-0.40.0 → kolega_code-0.41.1}/CHANGELOG.md

@@ -6,6 +6,26 @@ This project uses GitHub Releases for detailed generated release notes. This fil
 
 ## Unreleased
 
+## 0.41.1 - 2026-09-30
+
+### Fixed
+
+- Discovery tips no longer reappear after `/clear` or `/reset`; the quick-tip
+  modal is offered once per launch. Dismissing a tip now keeps it dismissed for
+  the rest of the run, and clearing the thread no longer brings it back.
+
+## 0.41.0 - 2026-09-30
+
+### Added
+
+- **New model support.** OpenAI **GPT-6 Sol** (`gpt-6-sol`), **GPT-6 Luna**
+  (`gpt-6-luna`) and **GPT-6.1 Sol** (`gpt-6.1-sol`) on both the `openai` and
+  `openai_chatgpt` providers; Anthropic **Claude Opus 5.5** (`claude-opus-5-5`),
+  **Claude Sonnet 5.5** (`claude-sonnet-5-5`) and **Claude Fable 5.1**
+  (`claude-fable-5-1`); and Fireworks **Ember-1**
+  (`accounts/fireworks/models/ember-1`). Every model was exercised against its
+  live API, including the ChatGPT-subscription backend.
+
 ## 0.40.0 - 2026-09-18
 
 ### Added

{kolega_code-0.40.0 → kolega_code-0.41.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kolega-code
-Version: 0.40.0
+Version: 0.41.1
 Summary: Multi-agent coding in the terminal
 Project-URL: Homepage, https://kolega-ai.github.io/kolega-code/
 Project-URL: Documentation, https://kolega-ai.github.io/kolega-code/

{kolega_code-0.40.0 → kolega_code-0.41.1}/kolega_code/__init__.py

@@ -174,4 +174,4 @@ __all__ = [
     "parse_git_status_output",
 ]
 
-__version__ = "0.40.0"
+__version__ = "0.41.1"

{kolega_code-0.40.0 → kolega_code-0.41.1}/kolega_code/agent/__init__.py

@@ -51,4 +51,4 @@ __all__ = [
     "ToolExtension",
 ]
 
-__version__ = "0.40.0"
+__version__ = "0.41.1"

{kolega_code-0.40.0 → kolega_code-0.41.1}/kolega_code/cli/app.py

@@ -379,6 +379,9 @@ class KolegaCodeApp(
         self._onboarding_screen: Optional[tui_onboarding.OnboardingScreen] = None
         self._discovery_tip_screen: DiscoveryTip | None = None
         self._discovery_tips_ready: bool = False
+        # The tip is a startup-only, once-per-launch offer: once shown or once the
+        # thread is reset, it is never re-armed for this process.
+        self._discovery_tip_consumed: bool = False
         self._onboarding_skipped = False
         self._permission_lock = asyncio.Lock()
         self._persistence_lock = asyncio.Lock()
@@ -1651,7 +1654,7 @@ class KolegaCodeApp(
         self._push_fullscreen_modal(screen)
 
     async def _maybe_show_discovery_tip(self) -> None:
-        """Offer one tip per fresh thread, without covering setup or prompts."""
+        """Offer one tip per launch, without covering setup or prompts."""
         if not self._discovery_tips_ready:
             return
         startup = next((entry for entry in self.conversation_entries if entry.kind == "startup"), None)
@@ -1666,7 +1669,10 @@ class KolegaCodeApp(
                 self._discovery_tip_screen.action_close()
             return
         if (
-            not eligible
+            # The offer is spent for this launch: a thread reset rebuilds the startup
+            # entry but must never re-arm the tip.
+            self._discovery_tip_consumed
+            or not eligible
             or startup is None
             or startup.startup_tip_shown
             or self.config is None
@@ -1682,6 +1688,7 @@ class KolegaCodeApp(
         ):
             return
         startup.startup_tip_shown = True
+        self._discovery_tip_consumed = True
         self._discovery_tip_screen = DiscoveryTip()
         await self.push_screen(self._discovery_tip_screen, callback=self._on_discovery_tip_dismissed)
 
@@ -3144,6 +3151,9 @@ class KolegaCodeApp(
         self._turn_active = False
         self._restore_composer_placeholder()
         self._set_chat_enabled(self.agent is not None)
+        # A thread reset is not startup: the launch's one tip offer stays spent, so
+        # the rebuilt startup entry must not re-arm it.
+        self._discovery_tip_consumed = True
         self._ensure_startup_entry(render=False)
         self._add_conversation_entry(
             tui_state.ConversationEntry(kind="progress", content=messages.THREAD_RESET_MESSAGE, complete=True)

{kolega_code-0.40.0 → kolega_code-0.41.1}/kolega_code/cli/provider_registry.py

@@ -88,6 +88,9 @@ MODEL_LABELS: dict[str, str] = {
     "claude-sonnet-4-5-20250929": "Claude Sonnet 4.5",
     "claude-opus-4-5-20251101": "Claude Opus 4.5",
     "claude-haiku-4-5-20251001": "Claude Haiku 4.5",
+    "claude-fable-5-1": "Claude Fable 5.1",
+    "claude-opus-5-5": "Claude Opus 5.5",
+    "claude-sonnet-5-5": "Claude Sonnet 5.5",
     # OpenAI (shared labels across the API and ChatGPT-subscription providers)
     "gpt-6-astra": "GPT-6 Astra",
     "gpt-5.6-sol": "GPT-5.6 Sol",
@@ -97,6 +100,9 @@ MODEL_LABELS: dict[str, str] = {
     "gpt-5.4": "GPT-5.4",
     "gpt-5.4-mini": "GPT-5.4 Mini",
     "gpt-5.3-codex-spark": "GPT-5.3 Codex Spark",
+    "gpt-6-sol": "GPT-6 Sol",
+    "gpt-6-luna": "GPT-6 Luna",
+    "gpt-6.1-sol": "GPT-6.1 Sol",
     # Google
     "gemini-3.8-flash": "Gemini 3.8 Flash",
     "gemini-3.7-flash": "Gemini 3.7 Flash",
@@ -118,6 +124,7 @@ MODEL_LABELS: dict[str, str] = {
     "accounts/fireworks/models/deepseek-v4-flash": "DeepSeek V4 Flash",
     "accounts/fireworks/models/minimax-m3": "MiniMax M3",
     "accounts/fireworks/models/qwen3p7-plus": "Qwen 3.7 Plus",
+    "accounts/fireworks/models/ember-1": "Ember-1",
     # Together
     "moonshotai/Kimi-K2.7-Code": "Kimi K2.7 Code",
     "zai-org/GLM-5.1": "GLM-5.1",

{kolega_code-0.40.0 → kolega_code-0.41.1}/kolega_code/llm/specs/catalog/anthropic.py

@@ -112,4 +112,49 @@ ANTHROPIC_SPECS = {
         "default_temperature": 1.0,
         "supports_vision": True,
     },
+    # Claude 5.1/5.5 generation (Sept 2026). All three keep the 1M context and
+    # 128K output ceiling, ship adaptive thinking that cannot be turned off, and
+    # support the full low/medium/high/xhigh/max effort ladder. `high` is
+    # Anthropic's documented default except on Opus 5.5, which defaults to
+    # `medium`. Appended after the older entries on purpose: catalog insertion
+    # order drives the Settings provider-switch fallback.
+    ("anthropic", "claude-fable-5-1"): {
+        "context_length": 1000000,
+        "max_completion_tokens": 128000,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_temperature": False,
+        "supports_vision": True,
+        "thinking_effort": ThinkingEffortSpec(
+            options=("low", "medium", "high", "xhigh", "max"),
+            default="high",
+            mode="anthropic_adaptive_effort",
+        ),
+    },
+    ("anthropic", "claude-opus-5-5"): {
+        "context_length": 1000000,
+        "max_completion_tokens": 128000,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_temperature": False,
+        "supports_vision": True,
+        "thinking_effort": ThinkingEffortSpec(
+            options=("low", "medium", "high", "xhigh", "max"),
+            default="medium",
+            mode="anthropic_adaptive_effort",
+        ),
+    },
+    ("anthropic", "claude-sonnet-5-5"): {
+        "context_length": 1000000,
+        "max_completion_tokens": 128000,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_temperature": False,
+        "supports_vision": True,
+        "thinking_effort": ThinkingEffortSpec(
+            options=("low", "medium", "high", "xhigh", "max"),
+            default="high",
+            mode="anthropic_adaptive_effort",
+        ),
+    },
 }

{kolega_code-0.40.0 → kolega_code-0.41.1}/kolega_code/llm/specs/catalog/fireworks.py

@@ -106,4 +106,24 @@ FIREWORKS_SPECS = {
             mode="openai_reasoning_effort",
         ),
     },
+    # Ember-1 (Fireworks Research research preview, 2026-09-22): a post-trained
+    # Kimi K3 that emits ~40% shorter reasoning traces at K3 quality. 1M window /
+    # 131072 output ceiling, matching the K3 entry on the same serverless stack.
+    # Unlike K3's max-only entry, Ember-1 honors the full effort ladder: probed
+    # live 2026-09-30, all five values are accepted and `none` measurably
+    # disables reasoning (0 reasoning tokens vs >0 at every other level).
+    # Appended last on purpose: catalog insertion order drives the Settings
+    # provider-switch fallback.
+    ("fireworks", "accounts/fireworks/models/ember-1"): {
+        "context_length": 1048576,
+        "max_completion_tokens": 131072,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_vision": True,
+        "thinking_effort": ThinkingEffortSpec(
+            options=("none", "low", "medium", "high", "max"),
+            default="medium",
+            mode="openai_reasoning_effort",
+        ),
+    },
 }

{kolega_code-0.40.0 → kolega_code-0.41.1}/kolega_code/llm/specs/catalog/openai.py

@@ -116,6 +116,59 @@ OPENAI_SPECS = {
             mode="openai_responses_reasoning",
         ),
     },
+    # GPT-6 Sol (2026-09-22) and GPT-6 Luna (2026-09-22) are the cheaper/faster
+    # members of the GPT-6 generation. Both support the `none` reasoning effort
+    # (unlike Astra and GPT-6.1 Sol) and share Astra's 1,050,000-token window and
+    # 128K output ceiling. Appended after the older entries on purpose: catalog
+    # insertion order drives the Settings provider-switch fallback.
+    ("openai", "gpt-6-sol"): {
+        "context_length": 1050000,
+        "max_completion_tokens": 128000,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_temperature": False,
+        "supports_vision": True,
+        "supports_hosted_web_search": True,
+        "preferred_edit_protocol": "codex_apply_patch",
+        "thinking_effort": ThinkingEffortSpec(
+            options=("none", "low", "medium", "high", "xhigh", "max"),
+            default="medium",
+            mode="openai_responses_reasoning",
+        ),
+    },
+    ("openai", "gpt-6-luna"): {
+        "context_length": 1050000,
+        "max_completion_tokens": 128000,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_temperature": False,
+        "supports_vision": True,
+        "supports_hosted_web_search": True,
+        "preferred_edit_protocol": "codex_apply_patch",
+        "thinking_effort": ThinkingEffortSpec(
+            options=("none", "low", "medium", "high", "xhigh", "max"),
+            default="medium",
+            mode="openai_responses_reasoning",
+        ),
+    },
+    # GPT-6.1 Sol (2026-09-29) is the near-Astra refresh: same window and output
+    # ceiling, but it does NOT accept `none`/`minimal` reasoning (a request with
+    # `none` is rejected — use `low` instead).
+    ("openai", "gpt-6.1-sol"): {
+        "context_length": 1050000,
+        "max_completion_tokens": 128000,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_temperature": False,
+        "supports_vision": True,
+        "supports_hosted_web_search": True,
+        "preferred_edit_protocol": "codex_apply_patch",
+        "thinking_effort": ThinkingEffortSpec(
+            options=("low", "medium", "high", "xhigh", "max"),
+            default="medium",
+            mode="openai_responses_reasoning",
+        ),
+    },
     # Note: gpt-5.3-codex-spark is intentionally NOT on the API-key `openai`
     # provider — it's a Codex model that 404s on Chat Completions and is only
     # reachable through the ChatGPT-subscription backend (openai_chatgpt) below.

{kolega_code-0.40.0 → kolega_code-0.41.1}/kolega_code/llm/specs/catalog/openai_chatgpt.py

@@ -136,4 +136,54 @@ OPENAI_CHATGPT_SPECS = {
             mode="openai_responses_reasoning",
         ),
     },
+    # GPT-6 Sol / Luna / 6.1 Sol reach Codex and ChatGPT Work on the
+    # subscription backend (mirrored here so the /model picker matches the
+    # Codex model picker). Budgets stay at the conservative subscription value
+    # used by the rest of the GPT-6 family until the Codex backend's larger
+    # window is verified, as with gpt-6-astra.
+    ("openai_chatgpt", "gpt-6-sol"): {
+        "context_length": 400000,
+        "max_completion_tokens": 128000,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_temperature": False,
+        "supports_vision": True,
+        "supports_hosted_web_search": True,
+        "preferred_edit_protocol": "codex_apply_patch",
+        "thinking_effort": ThinkingEffortSpec(
+            options=("none", "low", "medium", "high", "xhigh", "max"),
+            default="medium",
+            mode="openai_responses_reasoning",
+        ),
+    },
+    ("openai_chatgpt", "gpt-6-luna"): {
+        "context_length": 400000,
+        "max_completion_tokens": 128000,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_temperature": False,
+        "supports_vision": True,
+        "supports_hosted_web_search": True,
+        "preferred_edit_protocol": "codex_apply_patch",
+        "thinking_effort": ThinkingEffortSpec(
+            options=("none", "low", "medium", "high", "xhigh", "max"),
+            default="medium",
+            mode="openai_responses_reasoning",
+        ),
+    },
+    ("openai_chatgpt", "gpt-6.1-sol"): {
+        "context_length": 400000,
+        "max_completion_tokens": 128000,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_temperature": False,
+        "supports_vision": True,
+        "supports_hosted_web_search": True,
+        "preferred_edit_protocol": "codex_apply_patch",
+        "thinking_effort": ThinkingEffortSpec(
+            options=("low", "medium", "high", "xhigh", "max"),
+            default="medium",
+            mode="openai_responses_reasoning",
+        ),
+    },
 }

{kolega_code-0.40.0 → kolega_code-0.41.1}/pyproject.toml

@@ -8,7 +8,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "kolega-code"
-version = "0.40.0"
+version = "0.41.1"
 description = "Multi-agent coding in the terminal"
 readme = "README.md"
 license = "BUSL-1.1"