langroid 0.18.0__tar.gz → 0.18.1__tar.gz

{langroid-0.18.0 → langroid-0.18.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: langroid
-Version: 0.18.0
+Version: 0.18.1
 Summary: Harness LLMs with Multi-Agent Programming
 License: MIT
 Author: Prasad Chalasani
@@ -244,6 +244,8 @@ teacher_task.run()
 <summary> <b>Click to expand</b></summary>
 
 - **Oct 2024:**
+  - **[0.18.0]** [LLMConfig.async_stream_quiet](https://langroid.github.io/langroid/notes/async-streaming/) flag to 
+    turn off LLM output in async + stream mode.
   - **[0.17.0]** XML-based tools, see [docs](https://langroid.github.io/langroid/notes/xml-tools/).
 - **Sep 2024:**
   - **[0.16.0](https://github.com/langroid/langroid/releases/tag/0.16.0)**  Support for OpenAI `o1-mini` and `o1-preview` models.

{langroid-0.18.0 → langroid-0.18.1}/README.md

@@ -135,6 +135,8 @@ teacher_task.run()
 <summary> <b>Click to expand</b></summary>
 
 - **Oct 2024:**
+  - **[0.18.0]** [LLMConfig.async_stream_quiet](https://langroid.github.io/langroid/notes/async-streaming/) flag to 
+    turn off LLM output in async + stream mode.
   - **[0.17.0]** XML-based tools, see [docs](https://langroid.github.io/langroid/notes/xml-tools/).
 - **Sep 2024:**
   - **[0.16.0](https://github.com/langroid/langroid/releases/tag/0.16.0)**  Support for OpenAI `o1-mini` and `o1-preview` models.

{langroid-0.18.0 → langroid-0.18.1}/langroid/agent/base.py

@@ -1316,7 +1316,7 @@ class Agent(ABC):
         """
         Convert result of a responder (agent_response or llm_response, or task.run()),
         or tool handler, or handle_message_fallback,
-        to a ChatDocument, to enabling handling by other
+        to a ChatDocument, to enable handling by other
         responders/tasks in a task loop possibly involving multiple agents.
 
         Args:

{langroid-0.18.0 → langroid-0.18.1}/langroid/agent/callbacks/chainlit.py

@@ -244,8 +244,11 @@ class ChainlitAgentCallbacks:
         agent.callbacks.show_error_message = self.show_error_message
         agent.callbacks.show_start_response = self.show_start_response
         self.config = config
-
         self.agent: lr.Agent = agent
+        if self.agent.llm is not None:
+            # We don't want to suppress LLM output in async + streaming,
+            # since we often use chainlit async callbacks to display LLM output
+            self.agent.llm.config.async_stream_quiet = False
         if msg is not None:
             self.show_first_user_message(msg)
 

{langroid-0.18.0 → langroid-0.18.1}/langroid/agent/special/doc_chat_agent.py

@@ -248,12 +248,6 @@ class DocChatAgent(ChatAgent):
     def ingest(self) -> None:
         """
         Chunk + embed + store docs specified by self.config.doc_paths
-
-        Returns:
-            dict with keys:
-                n_splits: number of splits
-                urls: list of urls
-                paths: list of file paths
         """
         if len(self.config.doc_paths) == 0:
             # we must be using a previously defined collection
@@ -1110,6 +1104,14 @@ class DocChatAgent(ChatAgent):
         Returns:
 
         """
+
+        if (
+            self.vecdb is None
+            or self.vecdb.config.collection_name
+            not in self.vecdb.list_collections(empty=False)
+        ):
+            return []
+
         # if we are using cross-encoder reranking or reciprocal rank fusion (RRF),
         # we can retrieve more docs during retrieval, and leave it to the cross-encoder
         # or RRF reranking to whittle down to self.config.parsing.n_similar_docs
@@ -1275,6 +1277,13 @@ class DocChatAgent(ChatAgent):
             List[Document]: list of relevant extracts
 
         """
+        if (
+            self.vecdb is None
+            or self.vecdb.config.collection_name
+            not in self.vecdb.list_collections(empty=False)
+        ):
+            return query, []
+
         if len(self.dialog) > 0 and not self.config.assistant_mode:
             # Regardless of whether we are in conversation mode or not,
             # for relevant doc/chunk extraction, we must convert the query

{langroid-0.18.0 → langroid-0.18.1}/langroid/agent/tools/orchestration.py

@@ -200,12 +200,13 @@ class DonePassTool(PassTool):
 
 
 class ForwardTool(PassTool):
-    """Tool for forwarding the received msg (ChatDocument) to another agent.
+    """Tool for forwarding the received msg (ChatDocument) to another agent or entity.
     Similar to PassTool, but with a specified recipient agent.
     """
 
     purpose: str = """
-    To forward the current message to an <agent>.
+    To forward the current message to an <agent>, where <agent> 
+    could be the name of an agent, or an entity such as "user", "llm".
     """
     request: str = "forward_tool"
     agent: str

{langroid-0.18.0 → langroid-0.18.1}/langroid/parsing/document_parser.py

@@ -81,13 +81,25 @@ def is_plain_text(path_or_bytes: str | bytes) -> bool:
     else:
         content = path_or_bytes[:1024]
     try:
+        # Use magic to detect the MIME type
+        import magic
+
+        mime_type = magic.from_buffer(content, mime=True)
+
+        # Check if the MIME type is not a text type
+        if not mime_type.startswith("text/"):
+            return False
+
         # Attempt to decode the content as UTF-8
         content = content[: find_last_full_char(content)]
 
-        _ = content.decode("utf-8")
-        # Additional checks can go here, e.g., to verify that the content
-        # doesn't contain too many unusual characters for it to be considered text
-        return True
+        try:
+            _ = content.decode("utf-8")
+            # Additional checks can go here, e.g., to verify that the content
+            # doesn't contain too many unusual characters for it to be considered text
+            return True
+        except UnicodeDecodeError:
+            return False
     except UnicodeDecodeError:
         # If decoding fails, it's likely not plain text (or not encoded in UTF-8)
         return False
@@ -123,7 +135,8 @@ class DocumentParser(Parser):
         Returns:
             DocumentParser: An instance of a DocumentParser subclass.
         """
-        if DocumentParser._document_type(source, doc_type) == DocumentType.PDF:
+        inferred_doc_type = DocumentParser._document_type(source, doc_type)
+        if inferred_doc_type == DocumentType.PDF:
             if config.pdf.library == "fitz":
                 return FitzPDFParser(source, config)
             elif config.pdf.library == "pypdf":
@@ -138,7 +151,7 @@ class DocumentParser(Parser):
                 raise ValueError(
                     f"Unsupported PDF library specified: {config.pdf.library}"
                 )
-        elif DocumentParser._document_type(source, doc_type) == DocumentType.DOCX:
+        elif inferred_doc_type == DocumentType.DOCX:
             if config.docx.library == "unstructured":
                 return UnstructuredDocxParser(source, config)
             elif config.docx.library == "python-docx":
@@ -147,7 +160,7 @@ class DocumentParser(Parser):
                 raise ValueError(
                     f"Unsupported DOCX library specified: {config.docx.library}"
                 )
-        elif DocumentParser._document_type(source, doc_type) == DocumentType.DOC:
+        elif inferred_doc_type == DocumentType.DOC:
             return UnstructuredDocParser(source, config)
         else:
             source_name = source if isinstance(source, str) else "bytes"

{langroid-0.18.0 → langroid-0.18.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langroid"
-version = "0.18.0"
+version = "0.18.1"
 description = "Harness LLMs with Multi-Agent Programming"
 authors = ["Prasad Chalasani <pchalasani@gmail.com>"]
 readme = "README.md"