langroid 0.17.1__py3-none-any.whl → 0.18.1__py3-none-any.whl

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/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/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/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/language_models/base.py

@@ -39,6 +39,10 @@ ToolTypes = Literal["function"]
 
 
 class LLMConfig(BaseSettings):
+    """
+    Common configuration for all language models.
+    """
+
     type: str = "openai"
     streamer: Optional[Callable[[Any], None]] = noop_fn
     api_base: str | None = None
@@ -48,6 +52,7 @@ class LLMConfig(BaseSettings):
     completion_model: str = ""
     temperature: float = 0.0
     chat_context_length: int = 8000
+    async_stream_quiet: bool = True  # suppress streaming output in async mode?
     completion_context_length: int = 8000
     max_output_tokens: int = 1024  # generate at most this many tokens
     # if input length + max_output_tokens > context length of model,
@@ -149,6 +154,10 @@ class OpenAIToolSpec(BaseModel):
 
 
 class LLMTokenUsage(BaseModel):
+    """
+    Usage of tokens by an LLM.
+    """
+
     prompt_tokens: int = 0
     completion_tokens: int = 0
     cost: float = 0.0
@@ -173,6 +182,10 @@ class LLMTokenUsage(BaseModel):
 
 
 class Role(str, Enum):
+    """
+    Possible roles for a message in a chat.
+    """
+
     USER = "user"
     SYSTEM = "system"
     ASSISTANT = "assistant"

langroid/language_models/openai_gpt.py

@@ -708,7 +708,7 @@ class OpenAIGPT(LanguageModel):
         event_args = ""
         event_fn_name = ""
         event_tool_deltas: Optional[List[Dict[str, Any]]] = None
-
+        silent = is_async and self.config.async_stream_quiet
         # The first two events in the stream of Azure OpenAI is useless.
         # In the 1st: choices list is empty, in the 2nd: the dict delta has null content
         if chat:
@@ -727,42 +727,40 @@ class OpenAIGPT(LanguageModel):
             event_text = choices[0]["text"]
         if event_text:
             completion += event_text
-            if not is_async:
+            if not silent:
                 sys.stdout.write(Colors().GREEN + event_text)
                 sys.stdout.flush()
                 self.config.streamer(event_text)
         if event_fn_name:
             function_name = event_fn_name
             has_function = True
-            if not is_async:
+            if not silent:
                 sys.stdout.write(Colors().GREEN + "FUNC: " + event_fn_name + ": ")
                 sys.stdout.flush()
                 self.config.streamer(event_fn_name)
 
         if event_args:
             function_args += event_args
-            if not is_async:
+            if not silent:
                 sys.stdout.write(Colors().GREEN + event_args)
                 sys.stdout.flush()
                 self.config.streamer(event_args)
 
-        if event_tool_deltas is not None:
-            # print out streaming tool calls
+        if event_tool_deltas is not None and not silent:
+            # print out streaming tool calls, if not async
             for td in event_tool_deltas:
                 if td["function"]["name"] is not None:
                     tool_fn_name = td["function"]["name"]
-                    if not is_async:
-                        sys.stdout.write(
-                            Colors().GREEN + "OAI-TOOL: " + tool_fn_name + ": "
-                        )
-                        sys.stdout.flush()
-                        self.config.streamer(tool_fn_name)
+                    sys.stdout.write(
+                        Colors().GREEN + "OAI-TOOL: " + tool_fn_name + ": "
+                    )
+                    sys.stdout.flush()
+                    self.config.streamer(tool_fn_name)
                 if td["function"]["arguments"] != "":
                     tool_fn_args = td["function"]["arguments"]
-                    if not is_async:
-                        sys.stdout.write(Colors().GREEN + tool_fn_args)
-                        sys.stdout.flush()
-                        self.config.streamer(tool_fn_args)
+                    sys.stdout.write(Colors().GREEN + tool_fn_args)
+                    sys.stdout.flush()
+                    self.config.streamer(tool_fn_args)
 
         # show this delta in the stream
         if choices[0].get("finish_reason", "") in [
@@ -872,6 +870,7 @@ class OpenAIGPT(LanguageModel):
                     completion=completion,
                     function_args=function_args,
                     function_name=function_name,
+                    is_async=True,
                 )
                 if is_break:
                     break

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-0.17.1.dist-info → langroid-0.18.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: langroid
-Version: 0.17.1
+Version: 0.18.1
 Summary: Harness LLMs with Multi-Agent Programming
 License: MIT
 Author: Prasad Chalasani
@@ -244,7 +244,9 @@ teacher_task.run()
 <summary> <b>Click to expand</b></summary>
 
 - **Oct 2024:**
-  - **[0.17.0]** XML-based tools, see [docs](https://langroid.github.io/langroid/tutorials/xml-tools/).
+  - **[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.
   - **[0.15.0](https://github.com/langroid/langroid/releases/tag/0.15.0)** Cerebras API support -- run llama-3.1 models hosted on Cerebras Cloud (very fast inference).

{langroid-0.17.1.dist-info → langroid-0.18.1.dist-info}/RECORD

@@ -1,16 +1,16 @@
 langroid/__init__.py,sha256=z_fCOLQJPOw3LLRPBlFB5-2HyCjpPgQa4m4iY5Fvb8Y,1800
 langroid/agent/__init__.py,sha256=ll0Cubd2DZ-fsCMl7e10hf9ZjFGKzphfBco396IKITY,786
-langroid/agent/base.py,sha256=hX_IGnEWcIhFGaLyhkZr6HWe_nQ4y6ERko-2H9mTkhM,64047
+langroid/agent/base.py,sha256=yFR9yrUWbuIOufgVDPOSjFDIP28JvGJWqvEu0HXEMX8,64045
 langroid/agent/batch.py,sha256=QZdlt1563hx4l3AXrCaGovE-PNG93M3DsvQAbDzdiS8,13705
 langroid/agent/callbacks/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-langroid/agent/callbacks/chainlit.py,sha256=Qedk1-CBCgo9PdaIa7AboLBFCTgAMg9q5nGmoqpZ378,22050
+langroid/agent/callbacks/chainlit.py,sha256=oiZYfPyfkUfFqhVFBiobN3hAJEaYn0FBSglvNE7t17Q,22302
 langroid/agent/chat_agent.py,sha256=m1fs-XsBzusRicwBETId3dpTQG8qEaS0yym_o4mfAI0,49931
 langroid/agent/chat_document.py,sha256=FZ_PkeKU5OVp1IUlMvspfqxIXzlyd7J_F32DSYrxQ7E,17651
 langroid/agent/helpers.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 langroid/agent/junk,sha256=LxfuuW7Cijsg0szAzT81OjWWv1PMNI-6w_-DspVIO2s,339
 langroid/agent/openai_assistant.py,sha256=2rjCZw45ysNBEGNzQM4uf0bTC4KkatGYAWcVcW4xcek,34337
 langroid/agent/special/__init__.py,sha256=gik_Xtm_zV7U9s30Mn8UX3Gyuy4jTjQe9zjiE3HWmEo,1273
-langroid/agent/special/doc_chat_agent.py,sha256=r1uPunYf2lQcqYQ4fsD8Q5gB9cZyf7cn0KPcR_CLtrU,59065
+langroid/agent/special/doc_chat_agent.py,sha256=xIqBOyLax_jMU0UevxqXf_aQUrRkW6MQUKpKnKvaqkQ,59281
 langroid/agent/special/lance_doc_chat_agent.py,sha256=s8xoRs0gGaFtDYFUSIRchsgDVbS5Q3C2b2mr3V1Fd-Q,10419
 langroid/agent/special/lance_rag/__init__.py,sha256=QTbs0IVE2ZgDg8JJy1zN97rUUg4uEPH7SLGctFNumk4,174
 langroid/agent/special/lance_rag/critic_agent.py,sha256=OtFuHthKQLkdVkvuZ2m0GNq1qOYLqHkm1pfLRFnSg5c,9548
@@ -40,7 +40,7 @@ langroid/agent/tools/duckduckgo_search_tool.py,sha256=NhsCaGZkdv28nja7yveAhSK_w6
 langroid/agent/tools/file_tools.py,sha256=GjPB5YDILucYapElnvvoYpGJuZQ25ecLs2REv7edPEo,7292
 langroid/agent/tools/google_search_tool.py,sha256=y7b-3FtgXf0lfF4AYxrZ3K5pH2dhidvibUOAGBE--WI,1456
 langroid/agent/tools/metaphor_search_tool.py,sha256=qj4gt453cLEX3EGW7nVzVu6X7LCdrwjSlcNY0qJW104,2489
-langroid/agent/tools/orchestration.py,sha256=afIt8Ah0dfrY2TQ7aNrkxUFLV288VTQ0L40Yse_-HeY,10785
+langroid/agent/tools/orchestration.py,sha256=u_iQDwD5L9JgE39pCemmIRG3MxyrBsw6KpQYf7pH-Ek,10881
 langroid/agent/tools/recipient_tool.py,sha256=0m2kQhYKTeGujAxhSPqH5z6hSAhVB_Dqour6uul2U30,9427
 langroid/agent/tools/retrieval_tool.py,sha256=2q2pfoYbZNfbWQ0McxrtmfF0ekGglIgRl-6uF26pa-E,871
 langroid/agent/tools/rewind_tool.py,sha256=XAXL3BpNhCmBGYq_qi_sZfHJuIw7NY2jp4wnojJ7WRs,5606
@@ -67,10 +67,10 @@ langroid/language_models/.chainlit/config.toml,sha256=1t5lHORGzc2E6dkaO9P15jYHu2
 langroid/language_models/.chainlit/translations/en-US.json,sha256=DAFz2HjOFFfboCStrUfKFg2BpplJPK_OOtixwF_GivY,9931
 langroid/language_models/__init__.py,sha256=1sUGobooTqq77XC7LxKsvME0RgSd5GGmeyrPo9SMh4U,940
 langroid/language_models/azure_openai.py,sha256=G4le3j4YLHV7IwgB2C37hO3MKijZ1KjynbYlEvpIF7Y,6214
-langroid/language_models/base.py,sha256=3DjxxVL7heuTLYRvNMjaIKS-eP8pMDvu8Gbkb6N3B0U,22286
+langroid/language_models/base.py,sha256=L02ToM7o1y5OlM23xYX8QbriRMVUodgwbnHodl3US28,22542
 langroid/language_models/config.py,sha256=9Q8wk5a7RQr8LGMT_0WkpjY8S4ywK06SalVRjXlfCiI,378
 langroid/language_models/mock_lm.py,sha256=HuiAvjHiCfffYF5xjFJUq945HVTW0QPbeUUctOnNCzQ,3868
-langroid/language_models/openai_gpt.py,sha256=haPcPe3CXAXpdU39rxGThObfv_YV0UivnL2hEdyyaHk,64810
+langroid/language_models/openai_gpt.py,sha256=AlHg-jbf7PBFcS1iqb5z--upokDcWS7RJ4qpAmovsVw,64822
 langroid/language_models/prompt_formatter/__init__.py,sha256=2-5cdE24XoFDhifOLl8yiscohil1ogbP1ECkYdBlBsk,372
 langroid/language_models/prompt_formatter/base.py,sha256=eDS1sgRNZVnoajwV_ZIha6cba5Dt8xjgzdRbPITwx3Q,1221
 langroid/language_models/prompt_formatter/hf_formatter.py,sha256=PVJppmjRvD-2DF-XNC6mE05vTZ9wbu37SmXwZBQhad0,5055
@@ -82,7 +82,7 @@ langroid/parsing/agent_chats.py,sha256=sbZRV9ujdM5QXvvuHVjIi2ysYSYlap-uqfMMUKulr
 langroid/parsing/code-parsing.md,sha256=--cyyNiSZSDlIwcjAV4-shKrSiRe2ytF3AdSoS_hD2g,3294
 langroid/parsing/code_parser.py,sha256=AOxb3xbYpTBPP3goOm5dKfJdh5hS_2BhLVCEkifWZN8,3796
 langroid/parsing/config.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-langroid/parsing/document_parser.py,sha256=WGnA5ADwMHliGJt6WW9rc4RiFXQcKU33b5zdPiGrtEY,24265
+langroid/parsing/document_parser.py,sha256=ZGxgG4ytnCIah4HWk3ZrK3EGAMXDjAzI9KaPEXCbMm0,24589
 langroid/parsing/image_text.py,sha256=sbLIQ5nHe2UnYUksBaQsmZGaX-X0qgEpPd7CEzi_z5M,910
 langroid/parsing/para_sentence_split.py,sha256=AJBzZojP3zpB-_IMiiHismhqcvkrVBQ3ZINoQyx_bE4,2000
 langroid/parsing/parse_json.py,sha256=aADo38bAHQhC8on4aWZZzVzSDy-dK35vRLZsFI2ewh8,4756
@@ -137,8 +137,8 @@ langroid/vector_store/meilisearch.py,sha256=6frB7GFWeWmeKzRfLZIvzRjllniZ1cYj3Hmh
 langroid/vector_store/momento.py,sha256=qR-zBF1RKVHQZPZQYW_7g-XpTwr46p8HJuYPCkfJbM4,10534
 langroid/vector_store/qdrant_cloud.py,sha256=3im4Mip0QXLkR6wiqVsjV1QvhSElfxdFSuDKddBDQ-4,188
 langroid/vector_store/qdrantdb.py,sha256=v88lqFkepADvlN6lByUj9I4NEKa9X9lWH16uTPPbYrE,17457
-pyproject.toml,sha256=dR9DMCoHjNyfJEKqnlAz-hxr3smiZmI2KvQRzip5W9o,7179
-langroid-0.17.1.dist-info/LICENSE,sha256=EgVbvA6VSYgUlvC3RvPKehSg7MFaxWDsFuzLOsPPfJg,1065
-langroid-0.17.1.dist-info/METADATA,sha256=LRVlfLjGecpCWm57L0QvP7nWOA3I3o6FayGaqYZ_F78,56325
-langroid-0.17.1.dist-info/WHEEL,sha256=FMvqSimYX_P7y0a7UY-_Mc83r5zkBZsCYPm7Lr0Bsq4,88
-langroid-0.17.1.dist-info/RECORD,,
+pyproject.toml,sha256=4WsW43aYs4SPCgfi1cEPVqaOQIHDZRuyUz-S8Ka9oAY,7179
+langroid-0.18.1.dist-info/LICENSE,sha256=EgVbvA6VSYgUlvC3RvPKehSg7MFaxWDsFuzLOsPPfJg,1065
+langroid-0.18.1.dist-info/METADATA,sha256=wamviAoQosjm8zl2aLANeU22nc0fsiosnPt8weJNsvs,56486
+langroid-0.18.1.dist-info/WHEEL,sha256=FMvqSimYX_P7y0a7UY-_Mc83r5zkBZsCYPm7Lr0Bsq4,88
+langroid-0.18.1.dist-info/RECORD,,

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "langroid"
-version = "0.17.1"
+version = "0.18.1"
 description = "Harness LLMs with Multi-Agent Programming"
 authors = ["Prasad Chalasani <pchalasani@gmail.com>"]
 readme = "README.md"