langroid 0.40.0__tar.gz → 0.41.1__tar.gz

{langroid-0.40.0 → langroid-0.41.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: langroid
-Version: 0.40.0
+Version: 0.41.1
 Summary: Harness LLMs with Multi-Agent Programming
 Author-email: Prasad Chalasani <pchalasani@gmail.com>
 License: MIT
@@ -15,6 +15,7 @@ Requires-Dist: colorlog<7.0.0,>=6.7.0
 Requires-Dist: docling<3.0.0,>=2.16.0
 Requires-Dist: docstring-parser<1.0,>=0.16
 Requires-Dist: duckduckgo-search<7.0.0,>=6.0.0
+Requires-Dist: exa-py>=1.8.7
 Requires-Dist: faker<19.0.0,>=18.9.0
 Requires-Dist: fakeredis<3.0.0,>=2.12.1
 Requires-Dist: fire<1.0.0,>=0.5.0
@@ -48,6 +49,7 @@ Requires-Dist: redis<6.0.0,>=5.0.1
 Requires-Dist: requests-oauthlib<2.0.0,>=1.3.1
 Requires-Dist: requests<3.0.0,>=2.31.0
 Requires-Dist: rich<14.0.0,>=13.3.4
+Requires-Dist: tavily-python>=0.5.0
 Requires-Dist: thefuzz<1.0.0,>=0.20.0
 Requires-Dist: tiktoken<1.0.0,>=0.7.0
 Requires-Dist: trafilatura<2.0.0,>=1.5.0
@@ -106,6 +108,8 @@ Provides-Extra: docling
 Requires-Dist: docling<3.0.0,>=2.16.0; extra == 'docling'
 Provides-Extra: docx
 Requires-Dist: python-docx<2.0.0,>=1.1.0; extra == 'docx'
+Provides-Extra: exa
+Requires-Dist: exa-py>=1.8.7; extra == 'exa'
 Provides-Extra: fastembed
 Requires-Dist: fastembed<0.4.0,>=0.3.1; extra == 'fastembed'
 Provides-Extra: google-generativeai
@@ -141,6 +145,8 @@ Requires-Dist: pymupdf4llm<0.1.0,>=0.0.17; extra == 'pdf-parsers'
 Requires-Dist: pymupdf<2.0.0,>=1.23.3; extra == 'pdf-parsers'
 Requires-Dist: pypdf>=5.1.0; extra == 'pdf-parsers'
 Requires-Dist: pytesseract<0.4.0,>=0.3.10; extra == 'pdf-parsers'
+Provides-Extra: pinecone
+Requires-Dist: pinecone-client>=5.0.1; extra == 'pinecone'
 Provides-Extra: postgres
 Requires-Dist: pgvector>=0.3.6; extra == 'postgres'
 Requires-Dist: psycopg2-binary>=2.9.10; extra == 'postgres'
@@ -154,6 +160,8 @@ Provides-Extra: sql
 Requires-Dist: psycopg2<3.0.0,>=2.9.7; extra == 'sql'
 Requires-Dist: pymysql<2.0.0,>=1.1.0; extra == 'sql'
 Requires-Dist: sqlalchemy<3.0.0,>=2.0.19; extra == 'sql'
+Provides-Extra: tavily
+Requires-Dist: tavily-python>=0.5.0; extra == 'tavily'
 Provides-Extra: transformers
 Requires-Dist: huggingface-hub<1.0.0,>=0.21.2; extra == 'transformers'
 Requires-Dist: torch<3.0.0,>=2.0.0; extra == 'transformers'
@@ -163,6 +171,7 @@ Requires-Dist: unstructured[docx,pdf,pptx]<1.0.0,>=0.16.15; extra == 'unstructur
 Provides-Extra: vecdbs
 Requires-Dist: chromadb<=0.4.23,>=0.4.21; extra == 'vecdbs'
 Requires-Dist: lancedb<0.9.0,>=0.8.2; extra == 'vecdbs'
+Requires-Dist: pinecone-client>=5.0.1; extra == 'vecdbs'
 Requires-Dist: pyarrow<16.0.0,>=15.0.0; extra == 'vecdbs'
 Requires-Dist: tantivy<0.22.0,>=0.21.0; extra == 'vecdbs'
 Requires-Dist: weaviate-client>=4.9.6; extra == 'vecdbs'

langroid-0.41.1/langroid/agent/tools/exa_search_tool.py

@@ -0,0 +1,68 @@
+"""
+A tool to trigger a Exa search for a given query,
+(https://docs.exa.ai/reference/getting-started)
+and return the top results with their titles, links, summaries.
+Since the tool is stateless (i.e. does not need
+access to agent state), it can be enabled for any agent, without having to define a
+special method inside the agent: `agent.enable_message(ExaSearchTool)`
+
+NOTE: To use this tool, you need to:
+
+* set the EXA_API_KEY environment variables in
+your `.env` file, e.g. `EXA_API_KEY=your_api_key_here`
+(Note as of 28 Jan 2023, Metaphor renamed to Exa, so you can also use
+`EXA_API_KEY=your_api_key_here`)
+
+* install langroid with the `exa-py` extra, e.g.
+`pip install langroid[exa]` or `uv pip install langroid[exa]`
+or `poetry add langroid[exa]`  or `uv add langroid[exa]`
+(it installs the `exa_py` package from pypi).
+
+For more information, please refer to the official docs:
+https://exa.ai/
+"""
+
+from typing import List, Tuple
+
+from langroid.agent.tool_message import ToolMessage
+from langroid.parsing.web_search import exa_search
+
+
+class ExaSearchTool(ToolMessage):
+    request: str = "exa_search"
+    purpose: str = """
+            To search the web and return up to <num_results> 
+            links relevant to the given <query>. When using this tool,
+            ONLY show the required JSON, DO NOT SAY ANYTHING ELSE.
+            Wait for the results of the web search, and then use them to
+            compose your response.
+            """
+    query: str
+    num_results: int
+
+    def handle(self) -> str:
+        """
+        Conducts a search using the exa API based on the provided query
+        and number of results by triggering a exa_search.
+
+        Returns:
+            str: A formatted string containing the titles, links, and
+                summaries of each search result, separated by two newlines.
+        """
+
+        search_results = exa_search(self.query, self.num_results)
+        # return Title, Link, Summary of each result, separated by two newlines
+        results_str = "\n\n".join(str(result) for result in search_results)
+        return f"""
+        BELOW ARE THE RESULTS FROM THE WEB SEARCH. USE THESE TO COMPOSE YOUR RESPONSE:
+        {results_str}
+        """
+
+    @classmethod
+    def examples(cls) -> List["ToolMessage" | Tuple[str, "ToolMessage"]]:
+        return [
+            cls(
+                query="When was the Llama2 Large Language Model (LLM) released?",
+                num_results=3,
+            ),
+        ]

langroid-0.41.1/langroid/agent/tools/tavily_search_tool.py

@@ -0,0 +1,50 @@
+"""
+A tool to trigger a Tavily search for a given query, and return the top results with
+their titles, links, summaries. Since the tool is stateless (i.e. does not need
+access to agent state), it can be enabled for any agent, without having to define a
+special method inside the agent: `agent.enable_message(TavilySearchTool)`
+"""
+
+from typing import List, Tuple
+
+from langroid.agent.tool_message import ToolMessage
+from langroid.parsing.web_search import tavily_search
+
+
+class TavilySearchTool(ToolMessage):
+    request: str = "tavily_search"
+    purpose: str = """
+            To search the web and return up to <num_results> 
+            links relevant to the given <query>. When using this tool,
+            ONLY show the required JSON, DO NOT SAY ANYTHING ELSE.
+            Wait for the results of the web search, and then use them to
+            compose your response.
+            """
+    query: str
+    num_results: int
+
+    def handle(self) -> str:
+        """
+        Conducts a search using Tavily based on the provided query
+        and number of results by triggering a tavily_search.
+
+        Returns:
+            str: A formatted string containing the titles, links, and
+                summaries of each search result, separated by two newlines.
+        """
+        search_results = tavily_search(self.query, self.num_results)
+        # return Title, Link, Summary of each result, separated by two newlines
+        results_str = "\n\n".join(str(result) for result in search_results)
+        return f"""
+        BELOW ARE THE RESULTS FROM THE WEB SEARCH. USE THESE TO COMPOSE YOUR RESPONSE:
+        {results_str}
+        """
+
+    @classmethod
+    def examples(cls) -> List["ToolMessage" | Tuple[str, "ToolMessage"]]:
+        return [
+            cls(
+                query="When was the Llama2 Large Language Model (LLM) released?",
+                num_results=3,
+            ),
+        ]

{langroid-0.40.0 → langroid-0.41.1}/langroid/language_models/model_info.py

@@ -320,8 +320,15 @@ MODEL_INFO: Dict[str, ModelInfo] = {
 }
 
 
-def get_model_info(model: str | ModelName) -> ModelInfo:
+def get_model_info(
+    model: str | ModelName,
+    fallback_model: str | ModelName = "",
+) -> ModelInfo:
     """Get model information by name or enum value"""
+    return _get_model_info(model) or _get_model_info(fallback_model) or ModelInfo()
+
+
+def _get_model_info(model: str | ModelName) -> ModelInfo | None:
     if isinstance(model, str):
-        return MODEL_INFO.get(model) or ModelInfo()
-    return MODEL_INFO.get(model.value) or ModelInfo()
+        return MODEL_INFO.get(model)
+    return MODEL_INFO.get(model.value)

{langroid-0.40.0 → langroid-0.41.1}/langroid/language_models/openai_gpt.py

@@ -468,7 +468,10 @@ class OpenAIGPT(LanguageModel):
             self.supports_strict_tools = self.api_base is None
             self.supports_json_schema = (
                 self.api_base is None
-                and get_model_info(self.config.chat_model).has_structured_output
+                and get_model_info(  # look for family/provider-specific then generic
+                    self.chat_model_orig,  # e.g. "gemini/gemini-2.0-flash"
+                    self.config.chat_model,  # e.g. "gemini-2.0-flash"
+                ).has_structured_output
             )
 
         if settings.chat_model != "":
@@ -620,7 +623,10 @@ class OpenAIGPT(LanguageModel):
     def supports_functions_or_tools(self) -> bool:
         return (
             self.is_openai_chat_model()
-            and get_model_info(self.config.chat_model).has_tools
+            and get_model_info(
+                self.chat_model_orig,
+                self.config.chat_model,
+            ).has_tools
         )
 
     def is_openai_completion_model(self) -> bool:
@@ -644,7 +650,10 @@ class OpenAIGPT(LanguageModel):
         """
         List of params that are not supported by the current model
         """
-        model_info = get_model_info(self.config.chat_model)
+        model_info = get_model_info(
+            self.chat_model_orig,
+            self.config.chat_model,
+        )
         unsupported = set(model_info.unsupported_params)
         for param, model_list in OpenAI_API_ParamInfo().params.items():
             if (
@@ -659,7 +668,10 @@ class OpenAIGPT(LanguageModel):
         Map of param name -> new name for specific models.
         Currently main troublemaker is o1* series.
         """
-        return get_model_info(self.config.chat_model).rename_params
+        return get_model_info(
+            self.chat_model_orig,
+            self.config.chat_model,
+        ).rename_params
 
     def chat_context_length(self) -> int:
         """
@@ -671,7 +683,12 @@ class OpenAIGPT(LanguageModel):
             if self.config.use_completion_for_chat
             else self.config.chat_model
         )
-        return get_model_info(model).context_length
+        orig_model = (
+            self.config.completion_model
+            if self.config.use_completion_for_chat
+            else self.chat_model_orig
+        )
+        return get_model_info(orig_model, model).context_length
 
     def completion_context_length(self) -> int:
         """
@@ -683,7 +700,12 @@ class OpenAIGPT(LanguageModel):
             if self.config.use_chat_for_completion
             else self.config.completion_model
         )
-        return get_model_info(model).context_length
+        orig_model = (
+            self.chat_model_orig
+            if self.config.use_chat_for_completion
+            else self.config.completion_model
+        )
+        return get_model_info(orig_model, model).context_length
 
     def chat_cost(self) -> Tuple[float, float]:
         """
@@ -709,7 +731,10 @@ class OpenAIGPT(LanguageModel):
         return (
             self.config.stream
             and settings.stream
-            and get_model_info(self.config.chat_model).allows_streaming
+            and get_model_info(
+                self.chat_model_orig,
+                self.config.chat_model,
+            ).allows_streaming
             and not settings.quiet
         )
 

{langroid-0.40.0 → langroid-0.41.1}/langroid/parsing/search.py

@@ -118,7 +118,7 @@ def preprocess_text(text: str) -> str:
         str: The preprocessed text.
     """
     # Ensure the NLTK resources are available
-    for resource in ["punkt", "wordnet", "stopwords"]:
+    for resource in ["tokenizers/punkt", "corpora/wordnet", "corpora/stopwords"]:
         download_nltk_resource(resource)
 
     # Lowercase the text

{langroid-0.40.0 → langroid-0.41.1}/langroid/parsing/utils.py

@@ -28,12 +28,13 @@ def download_nltk_resource(resource: str) -> None:
     try:
         nltk.data.find(resource)
     except LookupError:
-        nltk.download(resource, quiet=True)
+        model = resource.split("/")[-1]
+        nltk.download(model, quiet=True)
 
 
 # Download punkt_tab resource at module import
-download_nltk_resource("punkt_tab")
-download_nltk_resource("gutenberg")
+download_nltk_resource("tokenizers/punkt_tab")
+download_nltk_resource("corpora/gutenberg")
 
 T = TypeVar("T")
 

{langroid-0.40.0 → langroid-0.41.1}/langroid/parsing/web_search.py

@@ -16,6 +16,8 @@ from duckduckgo_search import DDGS
 from googleapiclient.discovery import Resource, build
 from requests.models import Response
 
+from langroid.exceptions import LangroidImportError
+
 
 class WebSearchResult:
     """
@@ -109,13 +111,7 @@ def metaphor_search(query: str, num_results: int = 5) -> List[WebSearchResult]:
     try:
         from metaphor_python import Metaphor
     except ImportError:
-        raise ImportError(
-            "You are attempting to use the `metaphor_python` library;"
-            "To use it, please install langroid with the `metaphor` extra, e.g. "
-            "`pip install langroid[metaphor]` or `poetry add langroid[metaphor]` "
-            "or `uv add langroid[metaphor]`"
-            "(it installs the `metaphor_python` package from pypi)."
-        )
+        raise LangroidImportError("metaphor-python", "metaphor")
 
     client = Metaphor(api_key=api_key)
 
@@ -130,6 +126,53 @@ def metaphor_search(query: str, num_results: int = 5) -> List[WebSearchResult]:
     ]
 
 
+def exa_search(query: str, num_results: int = 5) -> List[WebSearchResult]:
+    """
+    Method that makes an API call by Exa client that queries
+    the top num_results links that matches the query. Returns a list
+    of WebSearchResult objects.
+
+    Args:
+        query (str): The query body that users wants to make.
+        num_results (int): Number of top matching results that we want
+            to grab
+    """
+
+    load_dotenv()
+
+    api_key = os.getenv("EXA_API_KEY")
+    if not api_key:
+        raise ValueError(
+            """
+            EXA_API_KEY environment variables are not set. 
+            Please set one of them to your API key, and try again.
+            """
+        )
+
+    try:
+        from exa_py import Exa
+    except ImportError:
+        raise LangroidImportError("exa-py", "exa")
+
+    client = Exa(api_key=api_key)
+
+    response = client.search(
+        query=query,
+        num_results=num_results,
+    )
+    raw_results = response.results
+
+    return [
+        WebSearchResult(
+            title=result.title or "",
+            link=result.url,
+            max_content_length=3500,
+            max_summary_length=300,
+        )
+        for result in raw_results
+    ]
+
+
 def duckduckgo_search(query: str, num_results: int = 5) -> List[WebSearchResult]:
     """
     Method that makes an API call by DuckDuckGo client that queries
@@ -154,3 +197,44 @@ def duckduckgo_search(query: str, num_results: int = 5) -> List[WebSearchResult]
         )
         for result in search_results
     ]
+
+
+def tavily_search(query: str, num_results: int = 5) -> List[WebSearchResult]:
+    """
+    Method that makes an API call to Tavily API that queries
+    the top `num_results` links that match the query. Returns a list
+    of WebSearchResult objects.
+
+    Args:
+        query (str): The query body that users wants to make.
+        num_results (int): Number of top matching results that we want
+            to grab
+    """
+
+    load_dotenv()
+
+    api_key = os.getenv("TAVILY_API_KEY")
+    if not api_key:
+        raise ValueError(
+            "TAVILY_API_KEY environment variable is not set. "
+            "Please set it to your API key and try again."
+        )
+
+    try:
+        from tavily import TavilyClient
+    except ImportError:
+        raise LangroidImportError("tavily-python", "tavily")
+
+    client = TavilyClient(api_key=api_key)
+    response = client.search(query=query, max_results=num_results)
+    search_results = response["results"]
+
+    return [
+        WebSearchResult(
+            title=result["title"],
+            link=result["url"],
+            max_content_length=3500,
+            max_summary_length=300,
+        )
+        for result in search_results
+    ]

{langroid-0.40.0 → langroid-0.41.1}/langroid/vector_store/__init__.py

@@ -23,11 +23,7 @@ try:
     MeiliSearch
     MeiliSearchConfig
     __all__.extend(["meilisearch", "MeiliSearch", "MeiliSearchConfig"])
-except ImportError:
-    pass
 
-
-try:
     from . import lancedb
     from .lancedb import LanceDB, LanceDBConfig
 
@@ -35,10 +31,6 @@ try:
     LanceDB
     LanceDBConfig
     __all__.extend(["lancedb", "LanceDB", "LanceDBConfig"])
-except ImportError:
-    pass
-
-try:
     from . import chromadb
     from .chromadb import ChromaDBConfig, ChromaDB
 
@@ -46,10 +38,7 @@ try:
     ChromaDB
     ChromaDBConfig
     __all__.extend(["chromadb", "ChromaDBConfig", "ChromaDB"])
-except ImportError:
-    pass
 
-try:
     from . import postgres
     from .postgres import PostgresDB, PostgresDBConfig
 
@@ -57,6 +46,7 @@ try:
     PostgresDB
     PostgresDBConfig
     __all__.extend(["postgres", "PostgresDB", "PostgresDBConfig"])
+
     from . import weaviatedb
     from .weaviatedb import WeaviateDBConfig, WeaviateDB
 
@@ -64,5 +54,13 @@ try:
     WeaviateDB
     WeaviateDBConfig
     __all__.extend(["weaviatedb", "WeaviateDB", "WeaviateDBConfig"])
+
+    from . import pineconedb
+    from .pineconedb import PineconeDB, PineconeDBConfig
+
+    pineconedb
+    PineconeDB
+    PineconeDBConfig
+    __all__.extend(["pineconedb", "PineconeDB", "PineconeDBConfig"])
 except ImportError:
     pass

{langroid-0.40.0 → langroid-0.41.1}/langroid/vector_store/base.py

@@ -59,6 +59,7 @@ class VectorStore(ABC):
         from langroid.vector_store.lancedb import LanceDB, LanceDBConfig
         from langroid.vector_store.meilisearch import MeiliSearch, MeiliSearchConfig
         from langroid.vector_store.momento import MomentoVI, MomentoVIConfig
+        from langroid.vector_store.pineconedb import PineconeDB, PineconeDBConfig
         from langroid.vector_store.postgres import PostgresDB, PostgresDBConfig
         from langroid.vector_store.qdrantdb import QdrantDB, QdrantDBConfig
         from langroid.vector_store.weaviatedb import WeaviateDB, WeaviateDBConfig
@@ -77,6 +78,8 @@ class VectorStore(ABC):
             return PostgresDB(config)
         elif isinstance(config, WeaviateDBConfig):
             return WeaviateDB(config)
+        elif isinstance(config, PineconeDBConfig):
+            return PineconeDB(config)
 
         else:
             logger.warning(