langroid 0.36.0__py3-none-any.whl → 0.36.1__py3-none-any.whl

langroid/agent/special/doc_chat_agent.py

@@ -15,6 +15,7 @@ pip install "langroid[hf-embeddings]"
 """
 
 import logging
+import textwrap
 from collections import OrderedDict
 from functools import cache
 from typing import Any, Callable, Dict, List, Optional, Set, Tuple, no_type_check
@@ -81,7 +82,7 @@ You will be given various passages from these documents, and asked to answer que
 about them, or summarize them into coherent answers.
 """
 
-CHUNK_ENRICHMENT_DELIMITER = "<##-##-##>"
+CHUNK_ENRICHMENT_DELIMITER = "\n<##-##-##>"
 
 has_sentence_transformers = False
 try:
@@ -810,9 +811,11 @@ class DocChatAgent(ChatAgent):
         return "\n".join(
             [
                 f"""
-                [{i+1}]
+                -----[EXTRACT #{i+1}]----------
                 {content}
                 {source}
+                -----END OF EXTRACT------------
+                
                 """
                 for i, (content, source) in enumerate(zip(contents, sources))
             ]
@@ -949,12 +952,13 @@ class DocChatAgent(ChatAgent):
                     continue
 
                 # Combine original content with questions in a structured way
-                combined_content = f"""
-                {doc.content}
-                
+                combined_content = textwrap.dedent(
+                    f"""\
+                {doc.content}                
                 {enrichment_config.delimiter}
                 {enrichment}
-                """.strip()
+                """
+                )
 
                 new_doc = doc.copy(
                     update={
@@ -1440,7 +1444,7 @@ class DocChatAgent(ChatAgent):
         delimiter = self.config.chunk_enrichment_config.delimiter
         return [
             (
-                doc.copy(update={"content": doc.content.split(delimiter)[0].strip()})
+                doc.copy(update={"content": doc.content.split(delimiter)[0]})
                 if doc.content and getattr(doc.metadata, "has_enrichment", False)
                 else doc
             )

langroid/parsing/parser.py

@@ -267,9 +267,11 @@ class Parser:
                 # Truncate the chunk text at the punctuation mark
                 chunk_text = chunk_text[: last_punctuation + 1]
 
-            # Remove any newline characters and strip any leading or
-            # trailing whitespace
-            chunk_text_to_append = re.sub(r"\n{2,}", "\n", chunk_text).strip()
+            # Replace redundant (3 or more) newlines with 2 newlines to preser
+            # paragraph separation!
+            # But do NOT strip leading/trailing whitespace, to preserve formatting
+            # (e.g. code blocks, or in case we want to stitch chunks back together)
+            chunk_text_to_append = re.sub(r"\n{3,}", "\n\n", chunk_text)
 
             if len(chunk_text_to_append) > self.config.discard_chunk_chars:
                 # Append the chunk text to the list of chunks

langroid/parsing/utils.py

@@ -310,9 +310,9 @@ def extract_numbered_segments(s: str, specs: str) -> str:
         ]
 
         # If we extracted any segments from this paragraph,
-        # join them and append to results
+        # join them with ellipsis (...) and append to results.
         if extracted_segments:
-            extracted_paragraphs.append(" ".join(extracted_segments))
+            extracted_paragraphs.append("...".join(extracted_segments))
 
     return "\n\n".join(extracted_paragraphs)
 

langroid/utils/output/citations.py

@@ -17,25 +17,45 @@ def extract_markdown_references(md_string: str) -> list[int]:
     return sorted(set(int(match) for match in matches))
 
 
-def format_footnote_text(content: str, width: int = 80) -> str:
+def format_footnote_text(content: str, width: int = 0) -> str:
     """
-    Formats the content part of a footnote (i.e. not the first line that
-    appears right after the reference [^4])
-    It wraps the text so that no line is longer than the specified width and indents
-    lines as necessary for markdown footnotes.
+    Formats the content so that each original line is individually processed.
+    - If width=0, no wrapping is done (lines remain as is).
+    - If width>0, lines are wrapped to that width.
+    - Blank lines remain blank (with indentation).
+    - Everything is indented by 4 spaces (for markdown footnotes).
 
     Args:
         content (str): The text of the footnote to be formatted.
-        width (int): Maximum width of the text lines.
+        width (int): Maximum width of the text lines. If 0, lines are not wrapped.
 
     Returns:
         str: Properly formatted markdown footnote text.
     """
     import textwrap
 
-    # Wrap the text to the specified width
-    wrapped_lines = textwrap.wrap(content, width)
-    if len(wrapped_lines) == 0:
-        return ""
-    indent = "    "  # Indentation for markdown footnotes
-    return indent + ("\n" + indent).join(wrapped_lines)
+    indent = "    "  # 4 spaces for markdown footnotes
+    lines = content.split("\n")  # keep original line structure
+
+    output_lines = []
+    for line in lines:
+        # If the line is empty (or just spaces), keep it blank (but indented)
+        if not line.strip():
+            output_lines.append(indent)
+            continue
+
+        if width > 0:
+            # Wrap each non-empty line to the specified width
+            wrapped = textwrap.wrap(line, width=width)
+            if not wrapped:
+                # If textwrap gives nothing, add a blank (indented) line
+                output_lines.append(indent)
+            else:
+                for subline in wrapped:
+                    output_lines.append(indent + subline)
+        else:
+            # No wrapping: just indent the original line
+            output_lines.append(indent + line)
+
+    # Join them with newline so we preserve the paragraph/blank line structure
+    return "\n".join(output_lines)

langroid/vector_store/base.py

@@ -264,7 +264,7 @@ class VectorStore(ABC):
             metadata = copy.deepcopy(id2metadata[w[0]])
             metadata.window_ids = w
             document = Document(
-                content=" ".join([d.content for d in self.get_documents_by_ids(w)]),
+                content="".join([d.content for d in self.get_documents_by_ids(w)]),
                 metadata=metadata,
             )
             # make a fresh id since content is in general different

{langroid-0.36.0.dist-info → langroid-0.36.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langroid
-Version: 0.36.0
+Version: 0.36.1
 Summary: Harness LLMs with Multi-Agent Programming
 Author-email: Prasad Chalasani <pchalasani@gmail.com>
 License: MIT
@@ -292,20 +292,28 @@ teacher_task.run()
 <summary> <b>Click to expand</b></summary>
 
 - **Jan 2025:**
-  - [0.33.0](https://github.com/langroid/langroid/releases/tag/0.33.3) Move from Poetry to uv!
+  - [0.36.0](https://github.com/langroid/langroid/releases/tag/0.36.0): Weaviate vector-db support (thanks @abab-dev).
+  - [0.35.0](https://github.com/langroid/langroid/releases/tag/0.35.0): Capture/Stream reasoning content from 
+    Reasoning LLMs (e.g. DeepSeek, OpenAI o1) in addition to final answer.
+  - [0.34.0](https://github.com/langroid/langroid/releases/tag/0.34.0): DocChatAgent 
+    chunk enrichment to improve retrieval. (collaboration with @dfm88). 
+  - [0.33.0](https://github.com/langroid/langroid/releases/tag/0.33.3) Move from Poetry to uv! (thanks @abab-dev).
   - [0.32.0](https://github.com/langroid/langroid/releases/tag/0.32.0) DeepSeek v3 support.
 - **Dec 2024:**
   - [0.31.0](https://github.com/langroid/langroid/releases/tag/0.31.0) Azure OpenAI Embeddings
-  - [0.30.0](https://github.com/langroid/langroid/releases/tag/0.30.0) Llama-cpp embeddings.
-  - [0.29.0](https://github.com/langroid/langroid/releases/tag/0.29.0) Custom Azure OpenAI Client 
+  - [0.30.0](https://github.com/langroid/langroid/releases/tag/0.30.0) Llama-cpp embeddings (thanks @Kwigg).
+  - [0.29.0](https://github.com/langroid/langroid/releases/tag/0.29.0) Custom Azure OpenAI Client (thanks 
+    @johannestang).
   - [0.28.0](https://github.com/langroid/langroid/releases/tag/0.28.0) `ToolMessage`: `_handler` field to override 
-default handler method name in `request` field. 
+default handler method name in `request` field (thanks @alexagr).
   - [0.27.0](https://github.com/langroid/langroid/releases/tag/0.27.0) OpenRouter Support.
   - [0.26.0](https://github.com/langroid/langroid/releases/tag/0.26.0) Update to latest Chainlit.
-  - [0.25.0](https://github.com/langroid/langroid/releases/tag/0.25.0) True Async Methods for agent and user-response.
+  - [0.25.0](https://github.com/langroid/langroid/releases/tag/0.25.0) True Async Methods for agent and 
+    user-response (thanks @alexagr).
 - **Nov 2024:**
   - **[0.24.0](https://langroid.github.io/langroid/notes/structured-output/)**: 
      Enables support for `Agent`s with strict JSON schema output format on compatible LLMs and strict mode for the OpenAI tools API.
+    (thanks @nilspalumbo).
   - **[0.23.0](https://langroid.github.io/langroid/tutorials/local-llm-setup/#local-llms-hosted-on-glhfchat)**: 
       support for LLMs (e.g. `Qwen2.5-Coder-32b-Instruct`) hosted on glhf.chat 
   - **[0.22.0](https://langroid.github.io/langroid/notes/large-tool-results/)**: 

{langroid-0.36.0.dist-info → langroid-0.36.1.dist-info}/RECORD

@@ -14,7 +14,7 @@ langroid/agent/xml_tool_message.py,sha256=6SshYZJKIfi4mkE-gIoSwjkEYekQ8GwcSiCv7a
 langroid/agent/callbacks/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 langroid/agent/callbacks/chainlit.py,sha256=RH8qUXaZE5o2WQz3WJQ1SdFtASGlxWCA6_HYz_3meDQ,20822
 langroid/agent/special/__init__.py,sha256=gik_Xtm_zV7U9s30Mn8UX3Gyuy4jTjQe9zjiE3HWmEo,1273
-langroid/agent/special/doc_chat_agent.py,sha256=8OYJ7IRepdQ9GpoTX_dhw0NkENmTe9iQsJAN6JI-09c,64670
+langroid/agent/special/doc_chat_agent.py,sha256=PCpdaVocIWt6ftO5OfmI1l20abVbKUSZWdMcu9hJDD0,64816
 langroid/agent/special/lance_doc_chat_agent.py,sha256=s8xoRs0gGaFtDYFUSIRchsgDVbS5Q3C2b2mr3V1Fd-Q,10419
 langroid/agent/special/lance_tools.py,sha256=qS8x4wi8mrqfbYV2ztFzrcxyhHQ0ZWOc-zkYiH7awj0,2105
 langroid/agent/special/relevance_extractor_agent.py,sha256=zIx8GUdVo1aGW6ASla0NPQjYYIpmriK_TYMijqAx3F8,4796
@@ -81,7 +81,7 @@ langroid/parsing/code_parser.py,sha256=AOxb3xbYpTBPP3goOm5dKfJdh5hS_2BhLVCEkifWZ
 langroid/parsing/document_parser.py,sha256=9xUOyrVNBAS9cpCvCptr2XK4Kq47W574i8zzGEoXc3c,24933
 langroid/parsing/para_sentence_split.py,sha256=AJBzZojP3zpB-_IMiiHismhqcvkrVBQ3ZINoQyx_bE4,2000
 langroid/parsing/parse_json.py,sha256=aADo38bAHQhC8on4aWZZzVzSDy-dK35vRLZsFI2ewh8,4756
-langroid/parsing/parser.py,sha256=N0jr1Zl_f_rx-8YMmSQftPHquqSQfec-3s7JAhhEe6I,12032
+langroid/parsing/parser.py,sha256=Wq204V1wqLdXS8kZ5J81dU2jE8fjoRY9zzNFbzLFDIs,12205
 langroid/parsing/repo_loader.py,sha256=3GjvPJS6Vf5L6gV2zOU8s-Tf1oq_fZm-IB_RL_7CTsY,29373
 langroid/parsing/routing.py,sha256=-FcnlqldzL4ZoxuDwXjQPNHgBe9F9-F4R6q7b_z9CvI,1232
 langroid/parsing/search.py,sha256=0i_r0ESb5HEQfagA2g7_uMQyxYPADWVbdcN9ixZhS4E,8992
@@ -89,7 +89,7 @@ langroid/parsing/spider.py,sha256=hAVM6wxh1pQ0EN4tI5wMBtAjIk0T-xnpi-ZUzWybhos,32
 langroid/parsing/table_loader.py,sha256=qNM4obT_0Y4tjrxNBCNUYjKQ9oETCZ7FbolKBTcz-GM,3410
 langroid/parsing/url_loader.py,sha256=JK48KktLRDBfjrt4nsUfy92M6yGdEeicAqOum2MdULM,4656
 langroid/parsing/urls.py,sha256=XjpaV5onG7gKQ5iQeFTzHSw5P08Aqw0g-rMUu61lR6s,7988
-langroid/parsing/utils.py,sha256=kb9DlHaG1iQB-6JagH1C26SdCNNf8U-2XaXia4_dWCw,12726
+langroid/parsing/utils.py,sha256=YrV2GNL4EOBGknA4AClPGdJ4S5B31radrt-Ou8OAKoU,12749
 langroid/parsing/web_search.py,sha256=8rW8EI3tyHITaB2l9MT_6yLMeQfo8y-Ih-8N2v2uMpk,4931
 langroid/prompts/__init__.py,sha256=RW11vK6jiLPuaUh4GpeFvstti73gkm8_rDMtrbo2YsU,142
 langroid/prompts/dialog.py,sha256=SpfiSyofSgy2pwD1YboHR_yHO3LEEMbv6j2sm874jKo,331
@@ -111,18 +111,18 @@ langroid/utils/types.py,sha256=4GrOnU3HLWh-UwaUPp7LlB3V413q3K5OSzc0ggDoQ6A,2510
 langroid/utils/algorithms/__init__.py,sha256=WylYoZymA0fnzpB4vrsH_0n7WsoLhmuZq8qxsOCjUpM,41
 langroid/utils/algorithms/graph.py,sha256=JbdpPnUOhw4-D6O7ou101JLA3xPCD0Lr3qaPoFCaRfo,2866
 langroid/utils/output/__init__.py,sha256=7P0f--4IZneNsTxXY5fd6d6iW-CeVe-KSsl-87sbBPc,340
-langroid/utils/output/citations.py,sha256=PSY2cpti8W-ZGFMAgj1lYoEIZy0lsniLpCliMsVkXtc,1425
+langroid/utils/output/citations.py,sha256=mQhRXVN-uhmKd2z32UZQBE0adZGEaQJ7cVXLfkrcZJI,2221
 langroid/utils/output/printing.py,sha256=yzPJZN-8_jyOJmI9N_oLwEDfjMwVgk3IDiwnZ4eK_AE,2962
 langroid/utils/output/status.py,sha256=rzbE7mDJcgNNvdtylCseQcPGCGghtJvVq3lB-OPJ49E,1049
 langroid/vector_store/__init__.py,sha256=BcoOm1tG3y0EqjkIGmMOHkY9iTUhDHgyruknWDKgqIg,1214
-langroid/vector_store/base.py,sha256=c9slwOcSWCG0SFGDuPLAQF9vBLDb4Eg8uaUol27Jf9c,14209
+langroid/vector_store/base.py,sha256=suBanIt0iKEgnMnGdQOyWS58guG20Jyy-GK4DMMuYL0,14208
 langroid/vector_store/chromadb.py,sha256=9WXW9IoSnhOmGEtMruVhEtVWL_VO6NXnPIz-nzh0gIQ,8235
 langroid/vector_store/lancedb.py,sha256=b3_vWkTjG8mweZ7ZNlUD-NjmQP_rLBZfyKWcxt2vosA,14855
 langroid/vector_store/meilisearch.py,sha256=6frB7GFWeWmeKzRfLZIvzRjllniZ1cYj3HmhHQICXLs,11663
 langroid/vector_store/momento.py,sha256=UNHGT6jXuQtqY9f6MdqGU14bVnS0zHgIJUa30ULpUJo,10474
 langroid/vector_store/qdrantdb.py,sha256=HRLCt-FG8y4718omwpFaQZnWeYxPj0XCwS4tjokI1sU,18116
 langroid/vector_store/weaviatedb.py,sha256=Jxe-cp2PyZdQ9NQVNZJ-CnsYsNxgUBdfAOoLZQEN650,10602
-langroid-0.36.0.dist-info/METADATA,sha256=aDLzYdeo80UbUZB0oEv9Rc1WMgWvG_jQtrBZnI2y5Bg,59508
-langroid-0.36.0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-langroid-0.36.0.dist-info/licenses/LICENSE,sha256=EgVbvA6VSYgUlvC3RvPKehSg7MFaxWDsFuzLOsPPfJg,1065
-langroid-0.36.0.dist-info/RECORD,,
+langroid-0.36.1.dist-info/METADATA,sha256=OzErGoPlFwxWia7jrFUx4M9FolTjexpJbgpTfhwT9Nk,60103
+langroid-0.36.1.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+langroid-0.36.1.dist-info/licenses/LICENSE,sha256=EgVbvA6VSYgUlvC3RvPKehSg7MFaxWDsFuzLOsPPfJg,1065
+langroid-0.36.1.dist-info/RECORD,,