langroid 0.1.247__py3-none-any.whl → 0.1.249__py3-none-any.whl

langroid/agent/base.py

@@ -35,7 +35,7 @@ from langroid.language_models.base import (
     LLMTokenUsage,
     StreamingIfAllowed,
 )
-from langroid.language_models.openai_gpt import OpenAIGPTConfig
+from langroid.language_models.openai_gpt import OpenAIGPT, OpenAIGPTConfig
 from langroid.mytypes import Entity
 from langroid.parsing.parse_json import extract_top_level_json
 from langroid.parsing.parser import Parser, ParsingConfig
@@ -94,12 +94,12 @@ class Agent(ABC):
         self._indent = ""
         self.llm = LanguageModel.create(config.llm)
         self.vecdb = VectorStore.create(config.vecdb) if config.vecdb else None
-        # token_encoding_model is used to obtain the tokenizer,
-        # so we ensure that the tokenizer corresponding to the model is used.
         if config.parsing is not None and self.config.llm is not None:
-            config.parsing.token_encoding_model = (
-                self.config.llm.chat_model or self.config.llm.completion_model
-            )
+            # token_encoding_model is used to obtain the tokenizer,
+            # so in case it's an OpenAI model, we ensure that the tokenizer
+            # corresponding to the model is used.
+            if isinstance(self.llm, OpenAIGPT) and self.llm.is_openai_chat_model():
+                config.parsing.token_encoding_model = self.llm.config.chat_model
         self.parser: Optional[Parser] = (
             Parser(config.parsing) if config.parsing else None
         )

langroid/agent/callbacks/chainlit.py

@@ -98,7 +98,7 @@ async def make_llm_settings_widgets(
         [
             cl.input_widget.TextInput(
                 id="chat_model",
-                label="Model Name (Default GPT4-Turbo)",
+                label="Model Name (Default GPT-4o)",
                 initial="",
                 placeholder="E.g. ollama/mistral or " "local/localhost:8000/v1",
             ),
@@ -337,6 +337,7 @@ class ChainlitAgentCallbacks:
         content: str,
         is_tool: bool = False,
         cached: bool = False,
+        language: str | None = None,
     ) -> None:
         """Show non-streaming LLM response."""
         step = cl.Step(
@@ -344,7 +345,7 @@ class ChainlitAgentCallbacks:
             name=self._entity_name("llm", tool=is_tool, cached=cached),
             type="llm",
             parent_id=self._get_parent_id(),
-            language="json" if is_tool else None,
+            language=language or ("json" if is_tool else None),
         )
         self.last_step = step
         self.curr_step = None

langroid/agent/chat_agent.py

@@ -15,6 +15,7 @@ from langroid.agent.tool_message import ToolMessage
 from langroid.language_models.base import (
     LLMFunctionSpec,
     LLMMessage,
+    LLMResponse,
     Role,
     StreamingIfAllowed,
 )
@@ -681,8 +682,11 @@ class ChatAgent(Agent):
             streamer = self.callbacks.start_llm_stream()
         self.llm.config.streamer = streamer
         with ExitStack() as stack:  # for conditionally using rich spinner
-            if not self.llm.get_stream():
+            if not self.llm.get_stream() and not settings.quiet:
                 # show rich spinner only if not streaming!
+                # (Why? b/c the intent of showing a spinner is to "show progress",
+                # and we don't need to do that when streaming, since
+                # streaming output already shows progress.)
                 cm = status(
                     "LLM responding to messages...",
                     log_if_quiet=False,
@@ -702,35 +706,22 @@ class ChatAgent(Agent):
             self.callbacks.finish_llm_stream(
                 content=str(response),
                 is_tool=self.has_tool_message_attempt(
-                    ChatDocument.from_LLMResponse(response, displayed=True)
+                    ChatDocument.from_LLMResponse(response, displayed=True),
                 ),
             )
         self.llm.config.streamer = noop_fn
         if response.cached:
             self.callbacks.cancel_llm_stream()
-
-        if not self.llm.get_stream() or response.cached:
-            # We would have already displayed the msg "live" ONLY if
-            # streaming was enabled, AND we did not find a cached response.
-            # If we are here, it means the response has not yet been displayed.
-            cached = f"[red]{self.indent}(cached)[/red]" if response.cached else ""
-            if not settings.quiet:
-                print(cached + "[green]" + escape(str(response)))
-                self.callbacks.show_llm_response(
-                    content=str(response),
-                    is_tool=self.has_tool_message_attempt(
-                        ChatDocument.from_LLMResponse(response, displayed=True)
-                    ),
-                    cached=response.cached,
-                )
+        self._render_llm_response(response)
         self.update_token_usage(
-            response,
+            response,  # .usage attrib is updated!
             messages,
             self.llm.get_stream(),
             chat=True,
             print_response_stats=self.config.show_stats and not settings.quiet,
         )
-        return ChatDocument.from_LLMResponse(response, displayed=True)
+        chat_doc = ChatDocument.from_LLMResponse(response, displayed=True)
+        return chat_doc
 
     async def llm_response_messages_async(
         self, messages: List[LLMMessage], output_len: Optional[int] = None
@@ -764,35 +755,65 @@ class ChatAgent(Agent):
             self.callbacks.finish_llm_stream(
                 content=str(response),
                 is_tool=self.has_tool_message_attempt(
-                    ChatDocument.from_LLMResponse(response, displayed=True)
+                    ChatDocument.from_LLMResponse(response, displayed=True),
                 ),
             )
         self.llm.config.streamer = noop_fn
         if response.cached:
             self.callbacks.cancel_llm_stream()
-        if not self.llm.get_stream() or response.cached:
+        self._render_llm_response(response)
+        self.update_token_usage(
+            response,  # .usage attrib is updated!
+            messages,
+            self.llm.get_stream(),
+            chat=True,
+            print_response_stats=self.config.show_stats and not settings.quiet,
+        )
+        chat_doc = ChatDocument.from_LLMResponse(response, displayed=True)
+        return chat_doc
+
+    def _render_llm_response(
+        self, response: ChatDocument | LLMResponse, citation_only: bool = False
+    ) -> None:
+        is_cached = (
+            response.cached
+            if isinstance(response, LLMResponse)
+            else response.metadata.cached
+        )
+        if self.llm is None:
+            return
+        if not citation_only and (not self.llm.get_stream() or is_cached):
+            # We expect response to be LLMResponse in this context
+            if not isinstance(response, LLMResponse):
+                raise ValueError(
+                    "Expected response to be LLMResponse, but got "
+                    f"{type(response)} instead."
+                )
             # We would have already displayed the msg "live" ONLY if
             # streaming was enabled, AND we did not find a cached response.
             # If we are here, it means the response has not yet been displayed.
-            cached = f"[red]{self.indent}(cached)[/red]" if response.cached else ""
+            cached = f"[red]{self.indent}(cached)[/red]" if is_cached else ""
             if not settings.quiet:
                 print(cached + "[green]" + escape(str(response)))
                 self.callbacks.show_llm_response(
                     content=str(response),
                     is_tool=self.has_tool_message_attempt(
-                        ChatDocument.from_LLMResponse(response, displayed=True)
+                        ChatDocument.from_LLMResponse(response, displayed=True),
                     ),
-                    cached=response.cached,
+                    cached=is_cached,
                 )
-
-        self.update_token_usage(
-            response,
-            messages,
-            self.llm.get_stream(),
-            chat=True,
-            print_response_stats=self.config.show_stats and not settings.quiet,
-        )
-        return ChatDocument.from_LLMResponse(response, displayed=True)
+        if isinstance(response, LLMResponse):
+            # we are in the context immediately after an LLM responded,
+            # we won't have citations yet, so we're done
+            return
+        if response.metadata.has_citation and not settings.quiet:
+            print("[grey37]SOURCES:\n" + response.metadata.source + "[/grey37]")
+            self.callbacks.show_llm_response(
+                content=str(response.metadata.source),
+                is_tool=False,
+                cached=False,
+                language="text",
+            )
 
     def _llm_response_temp_context(self, message: str, prompt: str) -> ChatDocument:
         """

langroid/agent/chat_document.py

@@ -53,6 +53,7 @@ class ChatDocMetaData(DocMetaData):
     usage: Optional[LLMTokenUsage]
     cached: bool = False
     displayed: bool = False
+    has_citation: bool = False
     status: Optional[StatusCode] = None
 
 

langroid/agent/special/doc_chat_agent.py

@@ -14,7 +14,7 @@ pip install "langroid[hf-embeddings]"
 """
 
 import logging
-from contextlib import ExitStack
+import re
 from functools import cache
 from typing import Any, Dict, List, Optional, Set, Tuple, no_type_check
 
@@ -49,7 +49,6 @@ from langroid.parsing.urls import get_list_from_user, get_urls_paths_bytes_indic
 from langroid.parsing.utils import batched
 from langroid.prompts.prompts_config import PromptsConfig
 from langroid.prompts.templates import SUMMARY_ANSWER_PROMPT_GPT4
-from langroid.utils.configuration import settings
 from langroid.utils.constants import NO_ANSWER
 from langroid.utils.output import show_if_debug, status
 from langroid.utils.pydantic_utils import dataframe_to_documents, extract_fields
@@ -83,6 +82,23 @@ except ImportError:
     pass
 
 
+def extract_citations(text: str) -> List[int]:
+    # Find all patterns that match [[<numbers>]]
+    matches = re.findall(r"\[\[([\d,]+)\]\]", text)
+
+    # Initialize a set to hold distinct citation numbers
+    citations: Set[int] = set()
+
+    # Process each match
+    for match in matches:
+        # Split numbers by comma and convert to integers
+        numbers = match.split(",")
+        citations.update(int(number) for number in numbers)
+
+    # Return a sorted list of unique citations
+    return sorted(citations)
+
+
 class DocChatAgentConfig(ChatAgentConfig):
     system_message: str = DEFAULT_DOC_CHAT_SYSTEM_MESSAGE
     user_message: str = DEFAULT_DOC_CHAT_INSTRUCTIONS
@@ -658,8 +674,7 @@ class DocChatAgent(ChatAgent):
             query_str = query_str[1:] if query_str is not None else None
             if self.llm is None:
                 raise ValueError("LLM not set")
-            with StreamingIfAllowed(self.llm, self.llm.get_stream()):
-                response = super().llm_response(query_str)
+            response = super().llm_response(query_str)
             if query_str is not None:
                 self.update_dialog(
                     query_str, "" if response is None else response.content
@@ -676,6 +691,10 @@ class DocChatAgent(ChatAgent):
         else:
             self.callbacks.show_start_response(entity="llm")
             response = self.answer_from_docs(query_str)
+            # Citation details (if any) are NOT generated by LLM
+            # (We extract these from LLM's numerical citations),
+            # so render them here
+            self._render_llm_response(response, citation_only=True)
             return ChatDocument(
                 content=response.content,
                 metadata=ChatDocMetaData(
@@ -701,8 +720,7 @@ class DocChatAgent(ChatAgent):
             query_str = query_str[1:] if query_str is not None else None
             if self.llm is None:
                 raise ValueError("LLM not set")
-            with StreamingIfAllowed(self.llm, self.llm.get_stream()):
-                response = await super().llm_response_async(query_str)
+            response = await super().llm_response_async(query_str)
             if query_str is not None:
                 self.update_dialog(
                     query_str, "" if response is None else response.content
@@ -719,6 +737,7 @@ class DocChatAgent(ChatAgent):
         else:
             self.callbacks.show_start_response(entity="llm")
             response = self.answer_from_docs(query_str)
+            self._render_llm_response(response, citation_only=True)
             return ChatDocument(
                 content=response.content,
                 metadata=ChatDocMetaData(
@@ -742,10 +761,11 @@ class DocChatAgent(ChatAgent):
         return "\n".join(
             [
                 f"""
+                [{i+1}]
                 {content}
                 {source}
                 """
-                for (content, source) in zip(contents, sources)
+                for i, (content, source) in enumerate(zip(contents, sources))
             ]
         )
 
@@ -769,7 +789,7 @@ class DocChatAgent(ChatAgent):
         # Substitute Q and P into the templatized prompt
 
         final_prompt = self.config.summarize_prompt.format(
-            question=f"Question:{question}", extracts=passages_str
+            question=question, extracts=passages_str
         )
         show_if_debug(final_prompt, "SUMMARIZE_PROMPT= ")
 
@@ -788,24 +808,24 @@ class DocChatAgent(ChatAgent):
         final_answer = answer_doc.content.strip()
         show_if_debug(final_answer, "SUMMARIZE_RESPONSE= ")
 
-        if final_answer.startswith("SOURCE"):
-            # sometimes SOURCE may be shown first,
-            # in this case just use final_answer as-is for both content and source
-            content = final_answer
-            sources = final_answer
-        else:
-            parts = final_answer.split("SOURCE:", maxsplit=1)
-            if len(parts) > 1:
-                content = parts[0].strip()
-                sources = parts[1].strip()
-            else:
-                content = final_answer
-                sources = ""
+        citations = extract_citations(final_answer)
+
+        citations_str = ""
+        if len(citations) > 0:
+            # append [i] source, content for each citation
+            citations_str = "\n".join(
+                [
+                    f"[{c}] {passages[c-1].metadata.source}\n{passages[c-1].content}"
+                    for c in citations
+                ]
+            )
+
         return ChatDocument(
-            content=content,
+            content=final_answer,  # does not contain citations
             metadata=ChatDocMetaData(
-                source="SOURCE: " + sources,
+                source=citations_str,  # only the citations
                 sender=Entity.LLM,
+                has_citation=len(citations) > 0,
                 cached=getattr(answer_doc.metadata, "cached", False),
             ),
         )
@@ -1277,15 +1297,7 @@ class DocChatAgent(ChatAgent):
                 content="\n\n".join([e.content for e in extracts]),
                 metadata=ChatDocMetaData(**meta),
             )
-        with ExitStack() as stack:
-            # conditionally use Streaming or rich console context
-            cm = (
-                StreamingIfAllowed(self.llm)
-                if settings.stream
-                else (status("LLM Generating final answer..."))
-            )
-            stack.enter_context(cm)  # type: ignore
-            response = self.get_summary_answer(query, extracts)
+        response = self.get_summary_answer(query, extracts)
 
         self.update_dialog(query, response.content)
         self.response = response  # save last response

langroid/agent/special/lance_rag/query_planner_agent.py

@@ -58,7 +58,7 @@ class LanceQueryPlanAgentConfig(ChatAgentConfig):
     - a possibly REPHRASED QUERY to be answerable given the FILTER.
         Keep in mind that the ASSISTANT does NOT know anything about the FILTER fields,
         so the REPHRASED QUERY should NOT mention ANY FILTER fields.
-        The answer will answer based on documents whose CONTENTS match the QUERY, 
+        The assistant will answer based on documents whose CONTENTS match the QUERY, 
         possibly REPHRASED. 
     - an OPTIONAL SINGLE-LINE Pandas-dataframe calculation/aggregation string 
         that can be used to calculate the answer to the original query, 
@@ -70,6 +70,10 @@ class LanceQueryPlanAgentConfig(ChatAgentConfig):
         
         IMPORTANT: The DataFrame `df` in this calculation is the result of 
         applying the FILTER AND REPHRASED QUERY to the documents.
+        
+        WATCH OUT!! When deciding the dataframe calc, if any, CAREFULLY
+        note what the query is asking, and ensure that the result of your
+        dataframe calc expression would answer the query.                
     
     
     EXAMPLE:

langroid/agent/special/table_chat_agent.py

@@ -25,7 +25,7 @@ from langroid.agent.tool_message import ToolMessage
 from langroid.language_models.openai_gpt import OpenAIChatModel, OpenAIGPTConfig
 from langroid.parsing.table_loader import read_tabular_data
 from langroid.prompts.prompts_config import PromptsConfig
-from langroid.utils.constants import DONE
+from langroid.utils.constants import DONE, PASS
 from langroid.vector_store.base import VectorStoreConfig
 
 logger = logging.getLogger(__name__)
@@ -52,7 +52,8 @@ in your code, and try another way, or use `pandas_eval` to explore the dataframe
 before submitting your final code. 
 
 Once you have the answer to the question, possibly after a few steps,
-say {DONE} and show me the answer. If you receive an error message, 
+say {DONE} and PRESENT THE ANSWER TO ME; do not just say {DONE}.
+If you receive an error message, 
 try using the `pandas_eval` tool/function again with the corrected code. 
 
 VERY IMPORTANT: When using the `pandas_eval` tool/function, DO NOT EXPLAIN ANYTHING,
@@ -238,12 +239,22 @@ class TableChatAgent(ChatAgent):
     def handle_message_fallback(
         self, msg: str | ChatDocument
     ) -> str | ChatDocument | None:
-        """Handle scenario where LLM forgets to say DONE or
-        forgets to use pandas_eval"""
+        """Handle various LLM deviations"""
         if isinstance(msg, ChatDocument) and msg.metadata.sender == lr.Entity.LLM:
+            if msg.content.strip() == DONE and self.sent_expression:
+                # LLM sent an expression (i.e. used the `pandas_eval` tool)
+                # but upon receiving the results, simply said DONE without
+                # narrating the result as instructed.
+                return """
+                    You forgot to PRESENT the answer to the user's query
+                    based on the results from `pandas_eval` tool.
+                """
             if self.sent_expression:
-                return DONE
+                # LLM forgot to say DONE
+                self.sent_expression = False
+                return DONE + " " + PASS
             else:
+                # LLM forgot to use the `pandas_eval` tool
                 return """
                     You forgot to use the `pandas_eval` tool/function 
                     to find the answer.

langroid/language_models/base.py

@@ -475,10 +475,12 @@ class LanguageModel(ABC):
         history = collate_chat_history(chat_history)
 
         prompt = f"""
-        Given the conversationn below, and a follow-up question, rephrase the follow-up 
-        question as a standalone question.
+        Given the CHAT HISTORY below, and a follow-up QUESTION or SEARCH PHRASE,
+        rephrase the follow-up question/phrase as a STANDALONE QUESTION that
+        can be understood without the context of the chat history.
         
         Chat history: {history}
+        
         Follow-up question: {question} 
         """.strip()
         show_if_debug(prompt, "FOLLOWUP->STANDALONE-PROMPT= ")

langroid/language_models/openai_gpt.py

@@ -582,7 +582,7 @@ class OpenAIGPT(LanguageModel):
 
     def get_stream(self) -> bool:
         """Get streaming status"""
-        return self.config.stream
+        return self.config.stream and settings.stream
 
     @no_type_check
     def _process_stream_event(

langroid/parsing/parser.py

@@ -56,7 +56,10 @@ class ParsingConfig(BaseSettings):
 class Parser:
     def __init__(self, config: ParsingConfig):
         self.config = config
-        self.tokenizer = tiktoken.encoding_for_model(config.token_encoding_model)
+        try:
+            self.tokenizer = tiktoken.encoding_for_model(config.token_encoding_model)
+        except Exception:
+            self.tokenizer = tiktoken.encoding_for_model("text-embedding-ada-002")
 
     def num_tokens(self, text: str) -> int:
         tokens = self.tokenizer.encode(text)

langroid/prompts/templates.py

@@ -50,26 +50,27 @@ EXTRACTION_PROMPT = f"""
 
 SUMMARY_ANSWER_PROMPT_GPT4 = f"""
 
-        Use the provided extracts (with sources)  to answer the question. 
+        Use the provided NUMBERED extracts (with sources)  to answer the QUESTION. 
         If there's not enough information, respond with {NO_ANSWER}. Use only the 
         information in these extracts, even if your answer is factually incorrect, 
         and even if the answer contradicts other parts of the document. The only 
         important thing is that your answer is consistent with and supported by the 
-        extracts. Compose your complete answer and cite all supporting sources on a 
-        separate separate line as follows. For each SOURCE, (a) show the source name,
-        and (b) the exact EXTRACT from the source that supports your answer.
-        DO NOT SHOW THE FULL EXTRACT; only show the FIRST 3 words and LAST 3 words.  
-        For example:
-        SOURCE:  https://en.wikipedia.org/wiki/Tree
-        EXTRACTS: "The tree species ... oak, maple, and birch."
+        extracts. Compose your complete answer, inserting CITATIONS 
+        in the format [[i,j,...]] where i,j,... are the extract NUMBERS you are citing.
+        For example your answer might look like this (NOTE HOW multiple citations
+        are grouped as [[2,5]]):
+        
+        Beethoven composed the 9th symphony in 1824. [[1]] After that he became deaf
+        and could not hear his own music. [[2,5]] He was a prolific composer and
+        wrote many famous pieces.
         
-        SOURCE: https://en.wikipedia.org/wiki/Oak
-        EXTRACTS: "The oak trees ... longevity and strength."
+        NUMBERED EXTRACTS:
         
         {{extracts}}
         
+        QUESTION:
         {{question}}
-        Answer:   
+
 """.strip()
 
 ANSWER_PROMPT_USE_HISTORY_GPT4 = f"""

langroid/utils/output/status.py

@@ -3,6 +3,7 @@ from contextlib import AbstractContextManager, ExitStack
 from typing import Any
 
 from rich.console import Console
+from rich.errors import LiveError
 
 from langroid.utils.configuration import quiet_mode, settings
 
@@ -19,15 +20,21 @@ def status(
     Displays a rich spinner if not in quiet mode, else optionally logs the message.
     """
     stack = ExitStack()
-
-    if settings.quiet:
-        if log_if_quiet:
-            logger.info(msg)
+    logged = False
     if settings.quiet and log_if_quiet:
+        logged = True
         logger.info(msg)
-    else:
-        stack.enter_context(console.status(msg))
 
+    if not settings.quiet:
+        try:
+            stack.enter_context(console.status(msg))
+        except LiveError:
+            if not logged:
+                logger.info(msg)
+
+    # When using rich spinner, we enforce quiet mode
+    # (since output will be messy otherwise);
+    # We make an exception to this when debug is enabled.
     stack.enter_context(quiet_mode(not settings.debug))
 
     return stack

{langroid-0.1.247.dist-info → langroid-0.1.249.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: langroid
-Version: 0.1.247
+Version: 0.1.249
 Summary: Harness LLMs with Multi-Agent Programming
 License: MIT
 Author: Prasad Chalasani

{langroid-0.1.247.dist-info → langroid-0.1.249.dist-info}/RECORD

@@ -1,21 +1,21 @@
 langroid/__init__.py,sha256=zsYpGiAUsvyzZzjm964NUamsJImrXSJPVGz9a2jE_uY,1679
 langroid/agent/__init__.py,sha256=_D8dxnfdr92ch1CIrUkKjrB5HVvsQdn62b1Fb2kBxV8,785
-langroid/agent/base.py,sha256=Aa8e_dHibXfpOXnldGCrVvObFKhO0P05Is97AuVGFLo,35569
+langroid/agent/base.py,sha256=5HQ9fAFTQL771rhx7UkVkL90bjxSVA1DjtvxV8_1RJA,35652
 langroid/agent/batch.py,sha256=feRA_yRG768ElOQjrKEefcRv6Aefd_yY7qktuYUQDwc,10040
 langroid/agent/callbacks/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-langroid/agent/callbacks/chainlit.py,sha256=uIfdmoOKIfzBVeM2ce0V5SCfHjcvYyaz13dfjFXCEN4,21020
-langroid/agent/chat_agent.py,sha256=X5uVMm9qdw3j-FRf4hbN8k8ByaSdtQCTuU8olKE0sbs,38750
-langroid/agent/chat_document.py,sha256=1ug58JXV3PEGjzRU9vQPjdZbRtza73ZFZksNwWBPziQ,9285
+langroid/agent/callbacks/chainlit.py,sha256=LboE3zlLLzClKxpBkzHX4XU6fW4lNZW97zwwN97uuaU,21067
+langroid/agent/chat_agent.py,sha256=YwlRMWoK_7vEl9pxgMeUjQAgsukrnZjDGJR0WVomSuQ,39592
+langroid/agent/chat_document.py,sha256=uwCq53SHRyxQw6qyhjzPYuJG48VHBgOf2122Ew3fk6c,9316
 langroid/agent/helpers.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 langroid/agent/junk,sha256=LxfuuW7Cijsg0szAzT81OjWWv1PMNI-6w_-DspVIO2s,339
 langroid/agent/openai_assistant.py,sha256=kIVDI4r-xGvplLU5s0nShPVHs6Jq-wOsfWE0kcMhAdQ,33056
 langroid/agent/special/__init__.py,sha256=NG0JkB5y4K0bgnd9Q9UIvFExun3uTfVOWEVLVymff1M,1207
-langroid/agent/special/doc_chat_agent.py,sha256=LwWNb_1s5n9rOk9OpOFPuuY1VnVX5DjzQmPwBanKRrM,53763
+langroid/agent/special/doc_chat_agent.py,sha256=SBatLDoa2_Ju_Gk_El8FmlMekgHmgpBU1ihx26yFIvc,54008
 langroid/agent/special/lance_doc_chat_agent.py,sha256=USp0U3eTaJzwF_3bdqE7CedSLbaqAi2tm-VzygcyLaA,10175
 langroid/agent/special/lance_rag/__init__.py,sha256=QTbs0IVE2ZgDg8JJy1zN97rUUg4uEPH7SLGctFNumk4,174
 langroid/agent/special/lance_rag/critic_agent.py,sha256=OsOcpcU_AmU2MagpZ5X5yxFeXyteKN9QJMzJGqIITig,6871
 langroid/agent/special/lance_rag/lance_rag_task.py,sha256=l_HQgrYY-CX2FwIsS961aEF3bYog3GDYo98fj0C0mSk,2889
-langroid/agent/special/lance_rag/query_planner_agent.py,sha256=M2ry4PL_UkAzDNFvdl_Kqb6YtOnWqqA9Qz361AAb_b4,9584
+langroid/agent/special/lance_rag/query_planner_agent.py,sha256=U_2V8l3M44R3mX-El3wG1k-u2lTToU-HGfKvRkEWoEA,9816
 langroid/agent/special/lance_tools.py,sha256=btMwKdcT8RdwAjmzbtN1xxm3s1H7ipO9GSpUamryYx8,1456
 langroid/agent/special/neo4j/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 langroid/agent/special/neo4j/csv_kg_chat.py,sha256=koL3sKtHm3aRkLTiARs54ngrcU3lOR1WaLLc_i8rWOU,6374
@@ -31,7 +31,7 @@ langroid/agent/special/sql/utils/description_extractors.py,sha256=RZ2R3DmASxB1ij
 langroid/agent/special/sql/utils/populate_metadata.py,sha256=x2OMKfmIBnJESBG3qKt6gvr3H3L4ZQcoxHfNdWfHjZs,2987
 langroid/agent/special/sql/utils/system_message.py,sha256=qKLHkvQWRQodTtPLPxr1GSLUYUFASZU8x-ybV67cB68,1885
 langroid/agent/special/sql/utils/tools.py,sha256=6uB2424SLtmapui9ggcEr0ZTiB6_dL1-JRGgN8RK9Js,1332
-langroid/agent/special/table_chat_agent.py,sha256=v8gRYj5frhYgfCSzmwL1DMmvyC5Qb5z_Ira_fwld5vk,9024
+langroid/agent/special/table_chat_agent.py,sha256=xz4nWndTYTykET_oaveHcZUQ8IEpmA5yB8QGTXfOifw,9624
 langroid/agent/task.py,sha256=b_d46txohISETxXJoWpmIX0hinvt1wjHbK08LZRBEz8,54020
 langroid/agent/tool_message.py,sha256=2kPsQUwi3ZzINTUNj10huKnZLjLp5SXmefacTHx8QDc,8304
 langroid/agent/tools/__init__.py,sha256=q-maq3k2BXhPAU99G0H6-j_ozoRvx15I1RFpPVicQIU,304
@@ -60,10 +60,10 @@ langroid/embedding_models/protoc/embeddings_pb2_grpc.py,sha256=9dYQqkW3JPyBpSEje
 langroid/embedding_models/remote_embeds.py,sha256=6_kjXByVbqhY9cGwl9R83ZcYC2km-nGieNNAo1McHaY,5151
 langroid/language_models/__init__.py,sha256=5L9ndEEC8iLJHjDJmYFTnv6-2-3xsxWUMHcugR8IeDs,821
 langroid/language_models/azure_openai.py,sha256=ncRCbKooqLVOY-PWQUIo9C3yTuKEFbAwyngXT_M4P7k,5989
-langroid/language_models/base.py,sha256=B6dX43ZR65mIvjD95W4RcfpT-WpmiuEcstR3eMrr56Y,21029
+langroid/language_models/base.py,sha256=2JhacnbQ-DJzLbOaJqyZPnl867xyiz_W-ODiAlEcp98,21131
 langroid/language_models/config.py,sha256=5UF3DzO1a-Dfsc3vghE0XGq7g9t_xDsRCsuRiU4dgBg,366
 langroid/language_models/openai_assistants.py,sha256=9K-DEAL2aSWHeXj2hwCo2RAlK9_1oCPtqX2u1wISCj8,36
-langroid/language_models/openai_gpt.py,sha256=Re7n8Vtb-03HktyxewCD9-gcgXyuN5tWjyW_TgJwees,50648
+langroid/language_models/openai_gpt.py,sha256=5O-roChpUDZ6OmRTOkR9e2AHHZL8w12Yzs9Q_xviDZ0,50668
 langroid/language_models/prompt_formatter/__init__.py,sha256=9JXFF22QNMmbQV1q4nrIeQVTtA3Tx8tEZABLtLBdFyc,352
 langroid/language_models/prompt_formatter/base.py,sha256=eDS1sgRNZVnoajwV_ZIha6cba5Dt8xjgzdRbPITwx3Q,1221
 langroid/language_models/prompt_formatter/hf_formatter.py,sha256=TFL6ppmeQWnzr6CKQzRZFYY810zE1mr8DZnhw6i85ok,5217
@@ -79,7 +79,7 @@ langroid/parsing/document_parser.py,sha256=sfyNhkIdAQkufOHb-tTv63EF51pGNzrnAl0DN
 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=tgB_oatcrgt6L9ZplC-xBBXjLzL1gjSQf1L2_W5kwFA,4230
-langroid/parsing/parser.py,sha256=vE5j1LVDeFQPmLrXCWBfvuoPsjjvVIGHcsIWCBR8HDM,10617
+langroid/parsing/parser.py,sha256=2TT6YMgEe79Kz9bPIqI-1RIEK77V2H2gbpbX5DhNNrY,10743
 langroid/parsing/repo_loader.py,sha256=My5UIe-h1xr0I-6Icu0ZVwRHmGRRRW8SrJYMc9J1M9Q,29361
 langroid/parsing/search.py,sha256=plQtjarB9afGfJLB0CyPXPq3mM4m7kRsfd0_4brziEI,8846
 langroid/parsing/spider.py,sha256=w_mHR1B4KOmxsBLoVI8kMkMTEbwTzeK3ath9fOMJrTk,3043
@@ -93,7 +93,7 @@ langroid/prompts/__init__.py,sha256=B0vpJzIJlMR3mFRtoQwyALsFzBHvLp9f92acD8xJA_0,
 langroid/prompts/chat-gpt4-system-prompt.md,sha256=Q3uLCJTPQvmUkZN2XDnkBC7M2K3X0F3C3GIQBaFvYvw,5329
 langroid/prompts/dialog.py,sha256=SpfiSyofSgy2pwD1YboHR_yHO3LEEMbv6j2sm874jKo,331
 langroid/prompts/prompts_config.py,sha256=XRQHzod7KBnoKn3B_V878jZiqBA7rcn-CtGPkuAe_yM,131
-langroid/prompts/templates.py,sha256=4X-07tnmUQ8Z_zaWRQAUUyKiErGztp3tERujqnG8sGA,6369
+langroid/prompts/templates.py,sha256=NxMyPIhDjmL3pNXBaNLrIsebETPQHr6VG5NWO_93NeA,6303
 langroid/prompts/transforms.py,sha256=GsQo1klGxUy0fACh6j0lTblk6XEl2erRnhRWlN2M4-c,2706
 langroid/utils/__init__.py,sha256=ARx5To4Hsv1K5QAzK4uUqdEoB_iq5HK797vae1AcMBI,300
 langroid/utils/algorithms/__init__.py,sha256=WylYoZymA0fnzpB4vrsH_0n7WsoLhmuZq8qxsOCjUpM,41
@@ -107,7 +107,7 @@ langroid/utils/llms/strings.py,sha256=CSAX9Z6FQOLXOzbLMe_Opqtc3ruDAKTTk7cPqc6Blh
 langroid/utils/logging.py,sha256=R8TN-FqVpwZ4Ajgls9TDMthLvPpQd0QVNXK-PJDj1Z8,3917
 langroid/utils/output/__init__.py,sha256=4X8Hdo1SEm06NUnggMJrLtW8i1owdDQPrS7J08BaTec,341
 langroid/utils/output/printing.py,sha256=5EsYB1O4qKhocW19aebOUzK82RD9U5nygbY21yo8gfg,2872
-langroid/utils/output/status.py,sha256=VoSXmWDuddo1ipCzDAA6qlgffr5E4lSmBD0rIdNxxcs,774
+langroid/utils/output/status.py,sha256=rzbE7mDJcgNNvdtylCseQcPGCGghtJvVq3lB-OPJ49E,1049
 langroid/utils/pandas_utils.py,sha256=UctS986Jtl_MvU5rA7-GfrjEHXP7MNu8ePhepv0bTn0,755
 langroid/utils/pydantic_utils.py,sha256=yb-ghaQYL7EIYeiZ0tailvZvAuJZNF7UBXkd3z35OYc,21728
 langroid/utils/system.py,sha256=tWoEbzHzJ6ywdsoa9EwsQrZfGk2t7q87_zKNwau2C8s,4546
@@ -121,7 +121,7 @@ langroid/vector_store/meilisearch.py,sha256=d2huA9P-NoYRuAQ9ZeXJmMKr7ry8u90RUSR2
 langroid/vector_store/momento.py,sha256=9cui31TTrILid2KIzUpBkN2Ey3g_CZWOQVdaFsA4Ors,10045
 langroid/vector_store/qdrant_cloud.py,sha256=3im4Mip0QXLkR6wiqVsjV1QvhSElfxdFSuDKddBDQ-4,188
 langroid/vector_store/qdrantdb.py,sha256=sk5Qb2ZNbooi0rorsMuqIMokF7WADw6PJ0D6goM2XBw,16802
-langroid-0.1.247.dist-info/LICENSE,sha256=EgVbvA6VSYgUlvC3RvPKehSg7MFaxWDsFuzLOsPPfJg,1065
-langroid-0.1.247.dist-info/METADATA,sha256=aFpf1jK3j5mmhySliisY85x10w54ikl3BAFi9s9tZmA,49426
-langroid-0.1.247.dist-info/WHEEL,sha256=FMvqSimYX_P7y0a7UY-_Mc83r5zkBZsCYPm7Lr0Bsq4,88
-langroid-0.1.247.dist-info/RECORD,,
+langroid-0.1.249.dist-info/LICENSE,sha256=EgVbvA6VSYgUlvC3RvPKehSg7MFaxWDsFuzLOsPPfJg,1065
+langroid-0.1.249.dist-info/METADATA,sha256=1md_MzQhAHs9J7_OEEOfiL8C4N8GlWj5BBQGK4HrkGw,49426
+langroid-0.1.249.dist-info/WHEEL,sha256=FMvqSimYX_P7y0a7UY-_Mc83r5zkBZsCYPm7Lr0Bsq4,88
+langroid-0.1.249.dist-info/RECORD,,