lionagi 0.0.313__tar.gz → 0.0.315__tar.gz

{lionagi-0.0.313/lionagi.egg-info → lionagi-0.0.315}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: lionagi
-Version: 0.0.313
+Version: 0.0.315
 Summary: Towards automated general intelligence.
 Author: HaiyangLi
 Author-email: Haiyang Li <ocean@lionagi.ai>
@@ -43,39 +43,35 @@ To contribute, you need to make a fork first, and then make pull request from yo
   
 # LionAGI
 
-
-**LionAGI is undergoing major transformation.**
+**Powerful Intelligent Workflow Automation**
 
 It is an intelligent agentic workflow automation framework. It introduces advanced ML models into any existing workflows and data infrastructure.
 
-#### Currently, it can
 
-- generate inferences from almost all models including local
-- run most model inferencing in parallel
-- produce structured pydantic outputs with flexible usage
+### Currently, it can
+
+- interact with almost any models including local*
+- run interactions in parallel for most models (OpenRouter, OpenAI, Ollama, litellm...)
+- produce structured pydantic outputs with flexible usage\*\*
 - automate workflow via graph based agents
 - use advanced prompting techniques, i.e. ReAct (reason-action)
 - …
-  
-#### It aims to:
 
-- provide a centralized agent-managed framework for ML-powered tools coordination.
+### It aims to:
+
+- provide a centralized agent-managed framework for, "ML-powered tools coordination".
+- The ways of coordination and possible path can occur among nodes is what we also refers to as `workflow` (the concept of workflow is still in design).
 - such that, people can utilize intelligence to solve their problems in real life.
 - achieve the goal by dramatically lowering the barrier of entries for creating use-case/domain specific tools.
 
-#### Why?
 
-- To better the human experience for more.
-
----
-
-
-**Powerful Intelligent Workflow Automation**
+All notebooks should run, as of 0.0.313,
 
-LionAGI is an **intelligent agent framework**. Tailored for **big data analysis** in conjunction with advanced **machine learning** tools, designed for data-centric, production-level projects. Lionagi provides a set of robust tools, enabling flexible and rapid design of agentic workflow, for your own data.  
+\* if there are models on providers that have not been configured, you can do so by configuring your own AI providers, and endpoints. 
 
+\*\* Structured Input/Output, Graph based agent system, as well as more advanced prompting techniques are undergoing fast interations...
 
-## Why Automating Workflows?
+### Why Automating Workflows?
 
 Intelligent AI models such as [Large Language Model (LLM)](https://en.wikipedia.org/wiki/Large_language_model), introduced new possibilities of human-computer interaction. LLMs is drawing a lot of attention worldwide due to its “one model fits all”, and incredible performance. One way of using LLM is to use as search engine, however, this usage is complicated by the fact that LLMs [hallucinate](https://arxiv.org/abs/2311.05232).
 

{lionagi-0.0.313 → lionagi-0.0.315}/README.md

@@ -12,39 +12,35 @@ To contribute, you need to make a fork first, and then make pull request from yo
   
 # LionAGI
 
-
-**LionAGI is undergoing major transformation.**
+**Powerful Intelligent Workflow Automation**
 
 It is an intelligent agentic workflow automation framework. It introduces advanced ML models into any existing workflows and data infrastructure.
 
-#### Currently, it can
 
-- generate inferences from almost all models including local
-- run most model inferencing in parallel
-- produce structured pydantic outputs with flexible usage
+### Currently, it can
+
+- interact with almost any models including local*
+- run interactions in parallel for most models (OpenRouter, OpenAI, Ollama, litellm...)
+- produce structured pydantic outputs with flexible usage\*\*
 - automate workflow via graph based agents
 - use advanced prompting techniques, i.e. ReAct (reason-action)
 - …
-  
-#### It aims to:
 
-- provide a centralized agent-managed framework for ML-powered tools coordination.
+### It aims to:
+
+- provide a centralized agent-managed framework for, "ML-powered tools coordination".
+- The ways of coordination and possible path can occur among nodes is what we also refers to as `workflow` (the concept of workflow is still in design).
 - such that, people can utilize intelligence to solve their problems in real life.
 - achieve the goal by dramatically lowering the barrier of entries for creating use-case/domain specific tools.
 
-#### Why?
 
-- To better the human experience for more.
-
----
-
-
-**Powerful Intelligent Workflow Automation**
+All notebooks should run, as of 0.0.313,
 
-LionAGI is an **intelligent agent framework**. Tailored for **big data analysis** in conjunction with advanced **machine learning** tools, designed for data-centric, production-level projects. Lionagi provides a set of robust tools, enabling flexible and rapid design of agentic workflow, for your own data.  
+\* if there are models on providers that have not been configured, you can do so by configuring your own AI providers, and endpoints. 
 
+\*\* Structured Input/Output, Graph based agent system, as well as more advanced prompting techniques are undergoing fast interations...
 
-## Why Automating Workflows?
+### Why Automating Workflows?
 
 Intelligent AI models such as [Large Language Model (LLM)](https://en.wikipedia.org/wiki/Large_language_model), introduced new possibilities of human-computer interaction. LLMs is drawing a lot of attention worldwide due to its “one model fits all”, and incredible performance. One way of using LLM is to use as search engine, however, this usage is complicated by the fact that LLMs [hallucinate](https://arxiv.org/abs/2311.05232).
 

{lionagi-0.0.313 → lionagi-0.0.315}/lionagi/__init__.py

@@ -8,6 +8,9 @@ from dotenv import load_dotenv
 
 from .core import direct, Branch, Session, func_to_tool
 from .integrations.provider.services import Services
+from .integrations.chunker.chunk import chunk
+from .integrations.loader.load import load
+
 
 logger = logging.getLogger(__name__)
 logger.setLevel(logging.INFO)

{lionagi-0.0.313 → lionagi-0.0.315}/lionagi/core/direct/__init__.py

@@ -1,6 +1,7 @@
 from .predict import predict
 from .select import select
 from .score import score
+from .react import react
 from .vote import vote
 
-__all__ = ["predict", "select", "score", "vote"]
+__all__ = ["predict", "select", "score", "vote", "react"]

lionagi-0.0.315/lionagi/core/direct/cot.py

@@ -0,0 +1 @@
+# TODO: chain of thoughts

{lionagi-0.0.313 → lionagi-0.0.315}/lionagi/core/direct/predict.py

@@ -6,9 +6,10 @@ using a language model. It includes fields for the input sentence, number of sen
 confidence score, and reason for the prediction.
 """
 
-from pydantic import Field
 from lionagi.libs import func_call
-from ..prompt.prompt_template import ScoredTemplate
+from lionagi.integrations.bridge.pydantic_.pydantic_bridge import Field
+
+from ..prompt.scored_template import ScoredTemplate
 from ..branch import Branch
 
 

lionagi-0.0.315/lionagi/core/direct/react.py

@@ -0,0 +1,167 @@
+from lionagi.libs import func_call, convert, AsyncUtil
+
+from lionagi.integrations.bridge.pydantic_.pydantic_bridge import Field
+from ..prompt.action_template import ActionedTemplate
+from ..branch import Branch
+
+
+class ReactTemplate(ActionedTemplate):
+    template_name: str = "default_react"
+    sentence: str | list | dict = Field(
+        default_factory=str,
+        description="the given sentence(s) to reason and take actions on",
+    )
+
+    def __init__(
+        self,
+        sentence=None,
+        instruction=None,
+        confidence_score=False,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.sentence = sentence
+        self.task = f"Think step by step. Perform reasoning and prepare actions with given tools only.Instruction: {instruction}. Absolutely DO NOT MAKE UP FUNCTIONS !!!"
+
+        if confidence_score:
+            self.output_fields.append("confidence_score")
+
+
+async def _react(
+    sentence,
+    *,
+    instruction=None,
+    branch=None,
+    confidence_score=False,
+    retries=2,
+    delay=0.5,
+    backoff_factor=2,
+    default_value=None,
+    timeout=None,
+    branch_name=None,
+    system=None,
+    messages=None,
+    service=None,
+    sender=None,
+    llmconfig=None,
+    tools=None,
+    datalogger=None,
+    persist_path=None,
+    tool_manager=None,
+    return_branch=False,
+    **kwargs,
+):
+
+    if "temperature" not in kwargs:
+        kwargs["temperature"] = 0.1
+
+    instruction = instruction or ""
+
+    branch = branch or Branch(
+        name=branch_name,
+        system=system,
+        messages=messages,
+        service=service,
+        sender=sender,
+        llmconfig=llmconfig,
+        tools=tools,
+        datalogger=datalogger,
+        persist_path=persist_path,
+        tool_manager=tool_manager,
+    )
+
+    _template = ReactTemplate(
+        sentence=sentence,
+        instruction=instruction,
+        confidence_score=confidence_score,
+    )
+
+    await func_call.rcall(
+        branch.chat,
+        prompt_template=_template,
+        retries=retries,
+        delay=delay,
+        backoff_factor=backoff_factor,
+        default=default_value,
+        timeout=timeout,
+        **kwargs,
+    )
+
+    if _template.action_needed:
+        actions = _template.actions
+        tasks = [branch.tool_manager.invoke(i.values()) for i in actions]
+        results = await AsyncUtil.execute_tasks(*tasks)
+
+        a = []
+        for idx, item in enumerate(actions):
+            res = {
+                "function": item["function"],
+                "arguments": item["arguments"],
+                "output": results[idx],
+            }
+            branch.add_message(response=res)
+            a.append(res)
+
+        _template.__setattr__("action_response", a)
+
+    return (_template, branch) if return_branch else _template
+
+
+async def react(
+    sentence,
+    *,
+    instruction=None,
+    num_instances=1,
+    branch=None,
+    confidence_score=False,
+    retries=2,
+    delay=0.5,
+    backoff_factor=2,
+    default_value=None,
+    timeout=None,
+    branch_name=None,
+    system=None,
+    messages=None,
+    service=None,
+    sender=None,
+    llmconfig=None,
+    tools=None,
+    datalogger=None,
+    persist_path=None,
+    tool_manager=None,
+    return_branch=False,
+    **kwargs,
+):
+
+    async def _inner(i=0):
+        return await _react(
+            sentence=sentence,
+            instruction=instruction,
+            num_instances=num_instances,
+            branch=branch,
+            confidence_score=confidence_score,
+            retries=retries,
+            delay=delay,
+            backoff_factor=backoff_factor,
+            default_value=default_value,
+            timeout=timeout,
+            branch_name=branch_name,
+            system=system,
+            messages=messages,
+            service=service,
+            sender=sender,
+            llmconfig=llmconfig,
+            tools=tools,
+            datalogger=datalogger,
+            persist_path=persist_path,
+            tool_manager=tool_manager,
+            return_branch=return_branch,
+            **kwargs,
+        )
+
+    if num_instances == 1:
+        return await _inner()
+
+    elif num_instances > 1:
+        return await func_call.alcall(range(num_instances), _inner)

{lionagi-0.0.313 → lionagi-0.0.315}/lionagi/core/direct/score.py

@@ -12,7 +12,7 @@ ScoreTemplate class and a language model.
 from pydantic import Field
 import numpy as np
 from lionagi.libs import func_call, convert
-from ..prompt.prompt_template import ScoredTemplate
+from ..prompt.scored_template import ScoredTemplate
 from ..branch import Branch
 
 
@@ -183,6 +183,7 @@ async def _score(
 
 async def score(
     sentence,
+    *,
     num_instances=1,
     instruction=None,
     score_range=(1, 10),

{lionagi-0.0.313 → lionagi-0.0.315}/lionagi/core/direct/select.py

@@ -13,7 +13,7 @@ from enum import Enum
 from pydantic import Field
 
 from lionagi.libs import func_call, StringMatch
-from ..prompt.prompt_template import ScoredTemplate
+from ..prompt.scored_template import ScoredTemplate
 from ..branch import Branch
 
 
@@ -39,6 +39,7 @@ class SelectTemplate(ScoredTemplate):
     answer: Enum | str = Field(
         default_factory=str, description="selection from given choices"
     )
+    choices: list = Field(default_factory=list, description="the given choices")
 
     signature: str = "sentence -> answer"
 

lionagi-0.0.315/lionagi/core/direct/sentiment.py

@@ -0,0 +1 @@
+# TODO: sentiment analysis

{lionagi-0.0.313 → lionagi-0.0.315}/lionagi/core/messages/schema.py

@@ -173,7 +173,7 @@ class Instruction(BaseMessage):
 
         if output_fields:
             format_ = f"""
-            Follow the following response format.
+            MUST EXACTLY Follow the following response format. NO ADDITIONAL COMMENTS ALLOWED!
             ```json
             {output_fields}
             ```

lionagi-0.0.315/lionagi/core/prompt/action_template.py

@@ -0,0 +1,26 @@
+from typing import Any
+from lionagi.integrations.bridge.pydantic_.pydantic_bridge import Field
+
+from .scored_template import ScoredTemplate
+
+
+class ActionRequest: ...
+
+
+class ActionedTemplate(ScoredTemplate):
+
+    action_needed: bool | None = Field(
+        False, description="true if actions are needed else false"
+    )
+
+    actions: list[dict | ActionRequest | Any] | None = Field(
+        default_factory=list,
+        description="""provide The list of action(s) to take, each action in {"function": function_name, "arguments": {param1:..., param2:..., ...}}. Leave blank if no further actions are needed, you must use provided parameters for each action, DO NOT MAKE UP KWARG NAME!!!""",
+    )
+
+    answer: str | dict | Any | None = Field(
+        default_factory=str,
+        description="output answer to the questions asked if further actions are not needed, leave blank if an accurate answer cannot be provided from context during this step",
+    )
+
+    signature: str = "sentence -> reason, action_needed, actions, answer"

{lionagi-0.0.313 → lionagi-0.0.315}/lionagi/core/prompt/field_validator.py

@@ -6,7 +6,45 @@ including numeric, boolean, string, and enum. It also provides a dictionary `val
 maps data types to their corresponding validation functions.
 """
 
-from lionagi.libs import convert, StringMatch
+from lionagi.libs import convert, StringMatch, ParseUtil
+
+
+def _has_action_keys(dict_):
+    return list(dict_.keys()) >= ["function", "arguments"]
+
+
+def check_action_field(x, fix_=True, **kwargs):
+    if (
+        isinstance(x, list)
+        and convert.is_same_dtype(x, dict)
+        and all(_has_action_keys(y) for y in x)
+    ):
+        return x
+    try:
+        x = _fix_action_field(x, fix_)
+        return x
+    except Exception as e:
+        raise ValueError("Invalid action field type.") from e
+
+
+def _fix_action_field(x, discard_=True):
+    corrected = []
+    if isinstance(x, str):
+        x = ParseUtil.fuzzy_parse_json(x)
+
+    try:
+        x = convert.to_list(x)
+
+        for i in x:
+            i = convert.to_dict(i)
+            if _has_action_keys(i):
+                corrected.append(i)
+            elif not discard_:
+                raise ValueError(f"Invalid action field: {i}")
+    except Exception as e:
+        raise ValueError(f"Invalid action field: {e}") from e
+
+    return corrected
 
 
 def check_number_field(x, fix_=True, **kwargs):
@@ -236,4 +274,5 @@ validation_funcs = {
     "bool": check_bool_field,
     "str": check_str_field,
     "enum": check_enum_field,
+    "action": check_action_field,
 }

{lionagi-0.0.313 → lionagi-0.0.315}/lionagi/core/prompt/prompt_template.py

@@ -207,6 +207,10 @@ class PromptTemplate(BaseComponent):
             setattr(self, k, v_)
             return True
 
+        if "lionagi.core.prompt.action_template.actionrequest" in str_:
+            self.__setattr__(k, validation_funcs["action"](v))
+            return True
+
         elif "bool" in str_:
             self.__setattr__(k, validation_funcs["bool"](v, fix_=fix_, **kwargs))
             return True
@@ -227,48 +231,50 @@ class PromptTemplate(BaseComponent):
             if k not in kwargs:
                 kwargs = {k: {}}
 
-            try:
-                if (
-                    self.model_fields[k].json_schema_extra["choices"] is not None
-                    and "choices" in self.model_fields[k].json_schema_extra
+            if self._field_has_choices(k):
+                self.choices[k] = self.model_fields[k].json_schema_extra["choices"]
+                if self._validate_field(
+                    k, v, choices=self.choices[k], fix_=fix_, **kwargs[k]
                 ):
-                    self.choices[k] = self.model_fields[k].json_schema_extra["choices"]
-                    if self._validate_field(
-                        k, v, choices=self.choices[k], fix_=fix_, **kwargs[k]
-                    ):
-                        continue
-                    else:
-                        raise ValueError(f"{k} has no choices")
-
-            except Exception as e:
-                if self._validate_field(k, v, fix_=fix_, **kwargs[k]):
                     continue
                 else:
-                    raise ValueError(f"failed to validate field {k}") from e
+                    raise ValueError(f"{k} has no choices")
+
+            elif self._validate_field(k, v, fix_=fix_, **kwargs[k]):
+                continue
+            else:
+                raise ValueError(f"failed to validate field {k}")
+
+    def _field_has_choices(self, k):
+        try:
+            a = (
+                self.model_fields[k].json_schema_extra["choices"] is not None
+                and "choices" in self.model_fields[k].json_schema_extra
+            )
+            return a if isinstance(a, bool) else False
+        except Exception:
+            return False
 
     def _process_response(self, out_, fix_=True):
         kwargs = self.out_validation_kwargs.copy()
         for k, v in out_.items():
             if k not in kwargs:
                 kwargs = {k: {}}
-            try:
-                if (
-                    self.model_fields[k].json_schema_extra["choices"] is not None
-                    and "choices" in self.model_fields[k].json_schema_extra
+
+            if self._field_has_choices(k):
+                self.choices[k] = self.model_fields[k].json_schema_extra["choices"]
+                if self._validate_field(
+                    k, v, choices=self.choices[k], fix_=fix_, **kwargs[k]
                 ):
-                    self.choices[k] = self.model_fields[k].json_schema_extra["choices"]
-                    if self._validate_field(
-                        k, v, choices=self.choices[k], fix_=fix_, **kwargs[k]
-                    ):
-                        continue
-                    else:
-                        raise ValueError(f"{k} has no choices")
-
-            except Exception as e:
-                if self._validate_field(k, v, fix_=fix_, **kwargs[k]):
                     continue
                 else:
-                    raise ValueError(f"failed to validate field {k}") from e
+                    raise ValueError(f"{k} has no choices")
+
+            elif self._validate_field(k, v, fix_=fix_, **kwargs[k]):
+                continue
+
+            else:
+                raise ValueError(f"failed to validate field {k} with value {v}")
 
     @property
     def in_(self):
@@ -288,16 +294,6 @@ class PromptTemplate(BaseComponent):
         return self
 
 
-class ScoredTemplate(PromptTemplate):
-    confidence_score: float | None = Field(
-        -1,
-        description="a numeric score between 0 to 1 formatted in num:0.2f",
-    )
-    reason: str | None = Field(
-        default_factory=str, description="brief reason for the given output"
-    )
-
-
 # class Weather(PromptTemplate):
 #     sunny: bool = Field(True, description="true if the weather is sunny outside else false")
 #     rainy: bool = Field(False, description="true if it is raining outside else false")

lionagi-0.0.315/lionagi/core/prompt/scored_template.py

@@ -0,0 +1,13 @@
+from lionagi.integrations.bridge.pydantic_.pydantic_bridge import Field
+
+from .prompt_template import PromptTemplate
+
+
+class ScoredTemplate(PromptTemplate):
+    confidence_score: float | None = Field(
+        -1,
+        description="a numeric score between 0 to 1 formatted in num:0.2f",
+    )
+    reason: str | None = Field(
+        default_factory=str, description="brief reason for the given output"
+    )

lionagi-0.0.315/lionagi/core/tool/manual.py

@@ -0,0 +1 @@
+# TODO: tool manual, instruction on how to use the tool for LLM

{lionagi-0.0.313 → lionagi-0.0.315}/lionagi/core/tool/tool_manager.py

@@ -38,7 +38,7 @@ class ToolManager:
     def has_tools(self):
         return self.registry != {}
 
-    def _register_tool(self, tool: Tool) -> None:
+    def _register_tool(self, tool: Tool | Callable) -> None:
         """
         Registers a tool in the registry. Raises a TypeError if the object is not an instance of Tool.
 
@@ -48,6 +48,8 @@ class ToolManager:
         Raises:
                 TypeError: If the provided object is not an instance of Tool.
         """
+        if isinstance(tool, Callable):
+            tool = func_to_tool(tool)[0]
         if not isinstance(tool, Tool):
             raise TypeError("Please register a Tool object.")
         name = tool.schema_["function"]["name"]

{lionagi-0.0.313 → lionagi-0.0.315}/lionagi/integrations/bridge/llamaindex_/node_parser.py

@@ -29,19 +29,18 @@ def get_llama_index_node_parser(node_parser: Any):
     import llama_index.core.node_parser
 
     if not isinstance(node_parser, str) and not issubclass(node_parser, NodeParser):
-        raise TypeError(f"node_parser must be a string or NodeParser.")
+        raise TypeError("node_parser must be a string or NodeParser.")
 
     if isinstance(node_parser, str):
         if node_parser == "CodeSplitter":
             SysUtil.check_import("tree_sitter_languages")
 
         try:
-            parser = getattr(llama_index.core.node_parser, node_parser)
-            return parser
+            return getattr(llama_index.core.node_parser, node_parser)
         except Exception as e:
             raise AttributeError(
                 f"llama_index_core has no such attribute:" f" {node_parser}, Error: {e}"
-            )
+            ) from e
 
     elif isinstance(node_parser, NodeParser):
         return node_parser
@@ -75,10 +74,8 @@ def llama_index_parse_node(
         parser = get_llama_index_node_parser(node_parser)
         try:
             parser = parser(*parser_args, **parser_kwargs)
-        except:
+        except Exception:
             parser = parser.from_defaults(*parser_args, **parser_kwargs)
-        nodes = parser.get_nodes_from_documents(documents)
-        return nodes
-
+        return parser.get_nodes_from_documents(documents)
     except Exception as e:
-        raise ValueError(f"Failed to parse. Error: {e}")
+        raise ValueError(f"Failed to parse. Error: {e}") from e

lionagi-0.0.315/lionagi/integrations/bridge/pydantic_/pydantic_bridge.py

@@ -0,0 +1 @@
+from pydantic import BaseModel, Field, ValidationError, AliasChoices, field_serializer