ragbits-core 1.4.0.dev202512021005__tar.gz → 1.4.0.dev202512050236__tar.gz

{ragbits_core-1.4.0.dev202512021005 → ragbits_core-1.4.0.dev202512050236}/.gitignore

@@ -8,6 +8,7 @@ venv/
 .venv/
 __pycache__/
 **.egg-info/
+.deepeval/
 
 # Local cursor rules
 .cursor/rules/local/
@@ -109,3 +110,6 @@ node_modules/
 lazygit
 
 lazygit.tar.gz
+
+# chat conversation logs
+duet_conversation.log

{ragbits_core-1.4.0.dev202512021005 → ragbits_core-1.4.0.dev202512050236}/CHANGELOG.md

@@ -1,22 +1,21 @@
 # CHANGELOG
 
 ## Unreleased
-- Add Support for Thinking in agents (#837)
 
+- Add Support for Thinking in agents (#837)
+- Add support for confirmation requests in chat (#853)
 - Add name parameter and slightly refactor HuggingFace dataloder (#829)
 
 ## 1.3.0 (2025-09-11)
 
 - feat: improve app startup speed with lazy LiteLLM loading
 
-
 - Feat: added support for IVFFlat indexing and Halfvec datatype
 
 - Added Lazy loading of dependencies in local.py and during importing of LiteLLM
 - Add tool_choice parameter to LLM interface (#738)
 - Fix Prompt consumes same iterator twice leading to no data added to chat (#768)
 
-
 ## 1.2.2 (2025-08-08)
 
 - Fix: rendering iterator arguments in Prompt (#768)
@@ -97,6 +96,7 @@
 ## 0.16.0 (2025-04-29)
 
 ## 0.15.0 (2025-04-28)
+
 - Allow using sparse embeddings with Qdrant and local vector stores (#493)
 - Add support for sparse embeddings in the Pgvector Vector Store (#493)
 - Added secret key environment variable / generation for signatures across ragbits packages (#537)
@@ -112,6 +112,7 @@
 - modify LocalEmbedder to use sentence-transformers instead of torch (#508)
 
 ## 0.13.0 (2025-04-02)
+
 - Make the score in VectorStoreResult consistent (always bigger is better)
 - Add router option to LiteLLMEmbedder (#440)
 - Make LLM / Embedder APIs consistent (#463)
@@ -120,11 +121,13 @@
 - Cost and capabilities config for custom litellm models (#481)
 
 ## 0.12.0 (2025-03-25)
+
 - Allow Prompt class to accept the asynchronous response_parser. Change the signature of parse_response method.
 - Fix from_config for LiteLLM class (#441)
 - Fix Qdrant vector store serialization (#419)
 
 ## 0.11.0 (2025-03-25)
+
 - Add HybridSearchVectorStore which can aggregate results from multiple VectorStores (#412)
 
 ## 0.10.2 (2025-03-21)
@@ -205,7 +208,6 @@
 - Add streaming interface to LLMs (#188).
 - Better images support in Prompt abstractions (#201).
 
-
 ## 0.3.0 (2024-11-06)
 
 ### Added
@@ -245,4 +247,4 @@
 - LiteLLM integration.
 - ChromaDB integration.
 - Prompts lab.
-- Prompts autodiscovery.
+- Prompts autodiscovery.

{ragbits_core-1.4.0.dev202512021005 → ragbits_core-1.4.0.dev202512050236}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ragbits-core
-Version: 1.4.0.dev202512021005
+Version: 1.4.0.dev202512050236
 Summary: Building blocks for rapid development of GenAI applications
 Project-URL: Homepage, https://github.com/deepsense-ai/ragbits
 Project-URL: Bug Reports, https://github.com/deepsense-ai/ragbits/issues

{ragbits_core-1.4.0.dev202512021005 → ragbits_core-1.4.0.dev202512050236}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "ragbits-core"
-version = "1.4.0.dev202512021005"
+version = "1.4.0.dev202512050236"
 description = "Building blocks for rapid development of GenAI applications"
 readme = "README.md"
 requires-python = ">=3.10"

{ragbits_core-1.4.0.dev202512021005 → ragbits_core-1.4.0.dev202512050236}/src/ragbits/core/llms/litellm.py

@@ -102,6 +102,7 @@ class LiteLLM(LLM[LiteLLMOptions], LazyLiteLLM):
         self.use_structured_output = use_structured_output
         self.router = router
         self.custom_model_cost_config = custom_model_cost_config
+        self._cached_router: Router | None = None  # Cache for auto-created router
         if custom_model_cost_config:
             self._litellm.register_model(custom_model_cost_config)
 
@@ -434,7 +435,14 @@ class LiteLLM(LLM[LiteLLMOptions], LazyLiteLLM):
         stream: bool = False,
         stream_options: dict | None = None,
     ) -> "ModelResponse | CustomStreamWrapper":
-        entrypoint = self.router or self._create_router_from_self_and_options(options)
+        # Use provided router, or create/reuse cached router
+        if self.router:
+            entrypoint = self.router
+        else:
+            # Create router only once and cache it to avoid callback leak
+            if self._cached_router is None:
+                self._cached_router = self._create_router_from_self_and_options(options)
+            entrypoint = self._cached_router
 
         # Preprocess messages for Claude with reasoning enabled
         processed_conversation = self._preprocess_messages_for_claude(conversation, options)