kolega-code 0.48.0__tar.gz → 0.49.0__tar.gz

{kolega_code-0.48.0 → kolega_code-0.49.0}/CHANGELOG.md

@@ -6,6 +6,31 @@ This project uses GitHub Releases for detailed generated release notes. This fil
 
 ## Unreleased
 
+## 0.49.0 - 2026-11-07
+
+### Added
+
+- **Inkling 2 Small.** The `thinking_machines` provider (Tinker's
+  Anthropic-compatible endpoint) gains the `thinkingmachines/Inkling-2-Small`
+  route: natively multimodal (text + image), with 256K-token context and
+  reasoning on by default. The thinking effort picker offers `none`, `low`,
+  `medium`, `high`, `xhigh` and `max`, and pricing follows the published
+  serverless rates.
+
+### Changed
+
+- The catalogued context window for Inkling and Inkling Small is now 256K
+  tokens, matching the published Tinker serverless routes (previously listed
+  as 1M).
+
+### Fixed
+
+- Project memory can be enabled when Kolega Code is launched from your home
+  directory, or any folder that contains its state directory. Memory showed
+  "Backend unavailable" there and refused to turn on. A failed memory change in
+  Settings also no longer reports that previous settings could not be restored
+  when nothing had been changed.
+
 ## 0.48.0 - 2026-11-06
 
 ### Added

{kolega_code-0.48.0 → kolega_code-0.49.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kolega-code
-Version: 0.48.0
+Version: 0.49.0
 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-0.48.0 → kolega_code-0.49.0}/kolega_code/__init__.py

@@ -174,4 +174,4 @@ __all__ = [
     "parse_git_status_output",
 ]
 
-__version__ = "0.48.0"
+__version__ = "0.49.0"

{kolega_code-0.48.0 → kolega_code-0.49.0}/kolega_code/agent/__init__.py

@@ -51,4 +51,4 @@ __all__ = [
     "ToolExtension",
 ]
 
-__version__ = "0.48.0"
+__version__ = "0.49.0"

{kolega_code-0.48.0 → kolega_code-0.49.0}/kolega_code/cli/provider_registry.py

@@ -220,6 +220,7 @@ MODEL_LABELS: dict[str, str] = {
     # Thinking Machines
     "thinkingmachines/Inkling": "Inkling",
     "thinkingmachines/Inkling-Small": "Inkling Small",
+    "thinkingmachines/Inkling-2-Small": "Inkling 2 Small",
     "glm-5.3-abliterated": "GLM-5.3 Abliterated",
     "glm-5.3-flash-abliterated": "GLM-5.3 Flash Abliterated",
     "deepseek-v4.1-flash-abliterated": "DeepSeek V4.1 Flash Abliterated",

{kolega_code-0.48.0 → kolega_code-0.49.0}/kolega_code/cli/tui/settings_screen.py

@@ -1149,28 +1149,32 @@ class SettingsScreen(ModalScreen[None]):
         enabled_changed = desired_enabled != status.enabled
         if not backend_changed and not enabled_changed:
             return True
+        backend_saved = enabled_saved = False
         try:
             if desired_backend is not None and backend_changed:
                 await asyncio.to_thread(manager.select_backend, desired_backend)
+                backend_saved = True
             if enabled_changed:
                 await asyncio.to_thread(manager.set_enabled, desired_enabled)
+                enabled_saved = True
             await self.owner._refresh_agent_memory()
         except Exception as error:
             rollback_errors: list[str] = []
-            if enabled_changed:
+            if enabled_saved:
                 try:
                     await asyncio.to_thread(manager.set_enabled, status.enabled)
                 except Exception as rollback_error:
                     rollback_errors.append(str(rollback_error))
-            if backend_changed:
+            if backend_saved:
                 try:
                     await asyncio.to_thread(manager.select_backend, status.backend_id)
                 except Exception as rollback_error:
                     rollback_errors.append(str(rollback_error))
-            try:
-                await self.owner._refresh_agent_memory()
-            except Exception as rollback_error:
-                rollback_errors.append(str(rollback_error))
+            if enabled_saved or backend_saved:
+                try:
+                    await self.owner._refresh_agent_memory()
+                except Exception as rollback_error:
+                    rollback_errors.append(str(rollback_error))
             detail = f"Could not update project memory: {error}"
             if rollback_errors:
                 detail += " Previous memory settings could not be fully restored."

{kolega_code-0.48.0 → kolega_code-0.49.0}/kolega_code/llm/specs/catalog/thinking_machines.py

@@ -3,6 +3,7 @@ from kolega_code.llm.specs.types import ThinkingEffortSpec
 # Thinking Machines Lab models — reached through Tinker's Anthropic-compatible
 # endpoint (base https://tinker.thinkingmachines.dev/services/tinker-prod/anthropic/api,
 # the Anthropic SDK appends /v1/messages). Model IDs are case-sensitive.
+# The published serverless routes use a 256K context window.
 #
 # Effort mirrors Tinker's documented set for the Messages endpoint: "low",
 # "medium", "high", "xhigh", "max" ("max" behaves the same as "xhigh"); "none"
@@ -10,7 +11,7 @@ from kolega_code.llm.specs.types import ThinkingEffortSpec
 # when effort is omitted is "high".
 THINKING_MACHINES_SPECS = {
     ("thinking_machines", "thinkingmachines/Inkling"): {
-        "context_length": 1000000,
+        "context_length": 262144,
         "max_completion_tokens": 32768,
         "input_budget": "window_minus_output",
         "default_temperature": 1.0,
@@ -53,7 +54,7 @@ THINKING_MACHINES_SPECS = {
         },
     },
     ("thinking_machines", "thinkingmachines/Inkling-Small"): {
-        "context_length": 1000000,
+        "context_length": 262144,
         "max_completion_tokens": 32768,
         "input_budget": "window_minus_output",
         "default_temperature": 1.0,
@@ -95,4 +96,47 @@ THINKING_MACHINES_SPECS = {
             ],
         },
     },
+    ("thinking_machines", "thinkingmachines/Inkling-2-Small"): {
+        "context_length": 262144,
+        "max_completion_tokens": 32768,
+        "input_budget": "window_minus_output",
+        "default_temperature": 1.0,
+        "supports_vision": True,
+        "preferred_edit_protocol": "claude_code",
+        "thinking_effort": ThinkingEffortSpec(
+            options=("none", "low", "medium", "high", "xhigh", "max"),
+            default="high",
+            mode="tinker_effort",
+        ),
+        "pricing": {
+            "schema_version": 1,
+            "pricing_id": "thinking_machines:thinkingmachines/Inkling-2-Small:2026-11-07",
+            "provider": "thinking_machines",
+            "model": "thinkingmachines/Inkling-2-Small",
+            "status": "priced",
+            "billing_mode": "serverless_inference",
+            "currency": "USD",
+            "unit": "per_million_tokens",
+            "rates": {"input": "0.30", "cache_read": "0.03", "output": "1.00"},
+            "checked_at": "2026-11-07",
+            "sources": [
+                {
+                    "url": "https://tinker-docs.thinkingmachines.ai/tinker/serverless.json",
+                    "model_id": "thinkingmachines/Inkling-2-Small:peft:262144:sampling-nvfp4",
+                    "checked_at": "2026-11-07",
+                }
+            ],
+            "conditions": {},
+            "schedules": [],
+            "exclusions": ["Taxes, account credits/free allowances, and negotiated discounts."],
+            "scope": "serverless_published_list_rates",
+            "cache_policy": {"cache_read": "separate", "cache_write": "included"},
+            "assumptions": [
+                "Published serverless rates for thinkingmachines/Inkling-2-Small:peft:262144:sampling-nvfp4 "
+                "are applied to the supported thinkingmachines/Inkling-2-Small alias as a catalogue estimate, "
+                "not a provider-reported charge.",
+                "Cache creation uses the published input rate; only cache reads have a separately published rate.",
+            ],
+        },
+    },
 }