lionagi 0.6.1__py3-none-any.whl → 0.7.0__py3-none-any.whl

lionagi/libs/token_transform/synthlang.py

@@ -0,0 +1,409 @@
+# synthlang.py
+"""
+SynthLang translator and system prompt generator.
+We incorporate the simpler framework approach (a list of frameworks to enable),
+and rely on a chosen template from the included FRAMEWORK_TEMPLATES.
+We also integrate optional perplexity compression from `compress_perplexity.py`.
+"""
+# For the optional compression step:
+from timeit import default_timer as timer
+from typing import Literal
+
+from lionagi.session.branch import Branch
+from lionagi.settings import Settings
+
+from .perplexity import compress_text, iModel
+
+########################################################################
+# 1) Template Data
+#    Each item is a dictionary capturing your multi-line text blocks.
+########################################################################
+
+group_theory = {
+    "title": "Group Theory Analysis",
+    "domain": "mathematics",
+    "category": "Abstract Algebra",
+    "overview": "Investigation of algebraic structures using group theory and ring theory.",
+    "content": """# Structures
+Let G be a group with operation •
+Let H be a subgroup of G
+Let N be a normal subgroup of G
+
+# Properties
+P(x): "x is a homomorphism"
+Q(x): "x preserves group structure"
+R(x): "x maps to kernel"
+
+# Objectives
+1. Prove fundamental homomorphism theorem
+2. Show group action properties
+3. Analyze quotient groups
+""",
+}
+
+category_theory = {
+    "title": "Category Theory Concepts",
+    "domain": "mathematics",
+    "category": "Category Theory",
+    "overview": "Abstract mathematical framework dealing with relationships between structures.",
+    "content": """# Basic Definitions
+C = (Ob(C), hom(C), ∘)
+F: C → D (Functor)
+...
+""",
+}
+
+complex_analysis = {
+    "title": "Complex Analysis Foundations",
+    "domain": "mathematics",
+    "category": "Complex Analysis",
+    "overview": "Advanced study of complex functions, including integration and residue theory.",
+    "content": """# Complex Integration
+∮ f(z)dz = 2πi∑Res(f,ak)
+...
+""",
+}
+
+math_logic = {
+    "title": "Mathematical Logic Foundations",
+    "domain": "mathematics",
+    "category": "Mathematical Logic",
+    "overview": "Exploration of fundamental mathematical logic concepts.",
+    "content": """# Sets and Axioms
+Let A be the set of all propositions.
+...
+""",
+}
+
+number_theory = {
+    "title": "Number Theory Principles",
+    "domain": "mathematics",
+    "category": "Number Theory",
+    "overview": "Study of integers, prime numbers, and arithmetic functions.",
+    "content": """# Congruence Relations
+a ≡ b (mod n)
+...
+""",
+}
+
+set_theory = {
+    "title": "Set Theory Foundations",
+    "domain": "mathematics",
+    "category": "Set Theory",
+    "overview": "Exploration of fundamental set theory concepts and operations.",
+    "content": """# Basic Definitions
+Let U be the universal set
+...
+""",
+}
+
+symbolic_systems = {
+    "title": "Symbolic Systems Analysis",
+    "domain": "computer-science",
+    "category": "Symbolic Systems",
+    "overview": "Analysis of symbolic computation in AI and formal languages.",
+    "content": """# Core Concepts
+Let Σ be a finite alphabet
+...
+""",
+}
+
+symbolic_exploration_system_supression = {
+    "title": "Systematic Suppression Exploration",
+    "domain": "societal",
+    "category": "Real-World Simulations",
+    "overview": "Analysis of suppression mechanisms using symbolic logic.",
+    "content": """# Sets and Categories
+Let U be the set of actions representing policies ...
+""",
+}
+
+topology_fundamentals = {
+    "title": "Topology Fundamentals",
+    "domain": "mathematics",
+    "category": "Topology",
+    "overview": "Study of geometric properties under continuous deformations.",
+    "content": """# Open Sets
+(X,τ) topological space
+...
+""",
+}
+
+FRAMEWORK_TEMPLATES = {
+    "group_theory": group_theory,
+    "category_theory": category_theory,
+    "complex_analysis": complex_analysis,
+    "math_logic": math_logic,
+    "number_theory": number_theory,
+    "set_theory": set_theory,
+    "symbolic_systems": symbolic_systems,
+    "symbolic_exploration_system_supression": symbolic_exploration_system_supression,
+    "topology_fundamentals": topology_fundamentals,
+}
+
+
+########################################################################
+# 2) Framework Definitions (Simplified)
+########################################################################
+
+# Instead of a complex dict with "selectedGlyphs", let's do a simpler approach:
+# We define each "framework" by name, description, and glyphs. The user can pick them by name.
+FRAMEWORK_OPTIONS = {
+    "math": {
+        "name": "Mathematical Framework",
+        "description": "Offers a suite of math glyphs and notation rules.",
+        "glyphs": [
+            {
+                "symbol": "↹",
+                "name": "Focus/Filter",
+                "description": "Used for focusing instructions",
+            },
+            {
+                "symbol": "Σ",
+                "name": "Summarize",
+                "description": "Condense large sets of data",
+            },
+            {
+                "symbol": "⊕",
+                "name": "Combine/Merge",
+                "description": "Merge multiple data sources",
+            },
+            {
+                "symbol": "•",
+                "name": "Group Operation",
+                "description": "Binary operation in group theory",
+            },
+        ],
+    },
+    "optim": {
+        "name": "Optimization Framework",
+        "description": "Compression and optimization for code/math expressions.",
+        "glyphs": [
+            {
+                "symbol": "IF",
+                "name": "Conditional Operator",
+                "description": "Represents branching logic",
+            }
+        ],
+    },
+    "custom_algebra": {
+        "name": "Custom Algebraic Framework",
+        "description": "Extra rules for ring, group, field expansions.",
+        "glyphs": [
+            {
+                "symbol": "∞",
+                "name": "Infinite Operator",
+                "description": "Represents unbounded algebraic ops",
+            }
+        ],
+    },
+}
+
+########################################################################
+# 3) SynthLang Base Prompt
+########################################################################
+
+BASE_PROMPT = """You are a SynthLang translator. You convert standard text into SynthLang format with these rules:
+
+[Framework Integration]
+- Use relevant framework glyphs wherever possible.
+- Combine multiple frameworks if requested.
+
+[Grammar Rules]
+1) Task Glyphs:
+   - ↹ (Focus/Filter) for main tasks
+   - Σ (Summarize) for condensing info
+   - ⊕ (Combine) for merging data
+   - ? (Query) for clarifications
+   - IF for conditionals
+
+2) Subject Markers:
+   - Use • before datasets (e.g., •myData)
+   - Use 花 for abstract concepts
+   - Use 山 for hierarchical structures
+
+3) Modifiers:
+   - ^format(type) for output format
+   - ^n for importance level
+   - ^lang for language
+   - ^t{n} for time constraints
+
+4) Flow Control:
+   - [p=n] for priority
+   - -> for sequential
+   - + for parallel
+   - | for alternatives
+
+[Output Style]
+- Maximize compression by removing articles & filler
+- Use glyphs for repeated phrases
+- Preserve semantic meaning
+"""
+
+
+########################################################################
+# 4) Utility to build the "Active Frameworks" block
+########################################################################
+
+
+def build_frameworks_text(frameworks: list[str]) -> str:
+    """
+    Takes a list of framework keys (e.g. ["math", "optim"]) and returns
+    a textual block describing them.
+    """
+    lines = []
+    if not frameworks:
+        frameworks = FRAMEWORK_OPTIONS.keys()
+
+    for fw_key in frameworks:
+        fw = FRAMEWORK_OPTIONS.get(fw_key, None)
+        if fw:
+            lines.append(f"{fw['name']}: {fw['description']}")
+            lines.append("Glyphs:")
+            for g in fw["glyphs"]:
+                lines.append(
+                    f"  {g['symbol']} -> {g['name']} ({g['description']})"
+                )
+            lines.append("")
+    return "\n".join(lines).strip()
+
+
+########################################################################
+# 5) Main: create the SynthLang prompt
+########################################################################
+
+
+def create_synthlang_system_prompt(
+    template_name: Literal[
+        "group_theory",
+        "category_theory",
+        "complex_analysis",
+        "math_logic",
+        "number_theory",
+        "set_theory",
+        "symbolic_systems",
+        "symbolic_exploration_system_supression",
+        "topology_fundamentals",
+    ],
+    frameworks: list[str],
+    additional_text: str = "",
+) -> str:
+    """
+    Merges:
+    - BASE_PROMPT
+    - "Active Frameworks" block (from frameworks list)
+    - The chosen template (from FRAMEWORK_TEMPLATES)
+    - plus additional text if desired
+    """
+    template_block = FRAMEWORK_TEMPLATES[template_name]
+    frameworks_text = build_frameworks_text(frameworks)
+
+    # We show the user the "domain", "category", etc. from the template
+    template_details = (
+        f"Title: {template_block['title']}\n"
+        f"Domain: {template_block['domain']}\n"
+        f"Category: {template_block['category']}\n"
+        f"Overview: {template_block['overview']}\n"
+        "Excerpt:\n"
+        f"{template_block['content']}\n"
+    )
+
+    prompt = (
+        f"{BASE_PROMPT}\n\n"
+        "[Active Frameworks]\n"
+        f"{frameworks_text}\n\n"
+        "[Template]\n"
+        f"{template_details}\n"
+    )
+
+    if additional_text.strip():
+        prompt += f"\n[Additional]\n{additional_text.strip()}\n"
+
+    # Usually you might want the entire final text to be the system instruction:
+    return prompt
+
+
+########################################################################
+# 6) Translate text into SynthLang format
+########################################################################
+
+
+# TODO: refactor using SynthLang package
+async def translate_to_synthlang(
+    text: str,
+    template_name: str = "symbolic_systems",
+    frameworks: list[str] = None,
+    compress: bool = False,
+    chat_model: iModel = None,
+    compress_model: iModel = None,
+    compression_ratio: float = 0.2,
+    compress_kwargs=None,
+    verbose: bool = True,
+    branch: Branch = None,
+    **kwargs,
+) -> str:
+    """
+    Optionally compress the input text using perplexity, then build a final
+    "SynthLang" style system prompt that includes frameworks, templates, etc.
+
+    :param text: The original text to be translated/compressed
+    :param template_name: Key in FRAMEWORK_TEMPLATES
+    :param frameworks: e.g. ["math", "optim"] to enable these frameworks
+    :param compress: If True, uses perplexity-based compression
+    :param chat_model: The model for perplexity compression
+    :param compression_ratio: float ratio of final token usage
+    :param compress_kwargs: Additional arguments to pass to compression
+    :return: A string in SynthLang style
+    """
+    from .synthlang import create_synthlang_system_prompt
+
+    start = timer()
+    chat_model = chat_model or iModel(**Settings.iModel.CHAT)
+    if compress:
+        text = await compress_text(
+            text,
+            chat_model=compress_model or chat_model,
+            target_ratio=compression_ratio,
+            **(compress_kwargs or {}),
+        )
+
+    # Build final system prompt
+    final_prompt = create_synthlang_system_prompt(
+        template_name=template_name,
+        frameworks=frameworks,
+    )
+
+    sys1 = None
+    sys2 = final_prompt
+    if branch and branch.system:
+        sys1 = branch.system
+        branch.msgs.add_message(system=sys2)
+
+    else:
+        branch = Branch(system=final_prompt, chat_model=chat_model)
+
+    from lionagi.service.endpoints.token_calculator import TokenCalculator
+
+    calculator = TokenCalculator()
+
+    len_tokens = calculator.tokenize(text, return_tokens=False)
+    out = await branch.communicate(
+        instruction=f"Converts the given text into SynthLang's hyper-efficient format.",
+        context="Text to convert:\n\n" + text,
+        guidance=f"Following SynthLang, translate the provided text into SynthLang syntax. Shrink the token size by 60-85%. Return only the translated text.",
+        **kwargs,
+    )
+    if sys1:
+        branch.msgs.add_message(system=sys1)
+
+    len_ = calculator.tokenize(out, return_tokens=False)
+    if verbose:
+        msg = ""
+        msg += f"Compression Method: SynthLang\n"
+        msg += f"Compressed Token number: {len_}\n"
+        msg += f"Token Compression Ratio: {len_ / len_tokens:.1%}\n"
+        msg += f"Compression Time: {timer() - start:.03f} seconds\n"
+        msg += f"Compression Model: {branch.chat_model.model_name}\n"
+        print(msg)
+
+    return out

lionagi/operations/ReAct/ReAct.py

@@ -0,0 +1,126 @@
+# Copyright (c) 2023 - 2024, HaiyangLi <quantocean.li at gmail dot com>
+#
+# SPDX-License-Identifier: Apache-2.0
+
+import logging
+from typing import TYPE_CHECKING, Any
+
+from pydantic import BaseModel
+
+from lionagi.operatives.types import Instruct
+from lionagi.service.imodel import iModel
+from lionagi.utils import copy
+
+if TYPE_CHECKING:
+    from lionagi.session.branch import Branch
+
+
+async def ReAct(
+    branch: "Branch",
+    instruct: Instruct | dict[str, Any],
+    interpret: bool = False,
+    tools: Any = None,
+    tool_schemas: Any = None,
+    response_format: type[BaseModel] = None,
+    extension_allowed: bool = False,
+    max_extensions: int | None = None,
+    response_kwargs: dict | None = None,
+    return_analysis: bool = False,
+    analysis_model: iModel | None = None,
+    **kwargs,
+):
+    # If no tools or tool schemas are provided, default to "all tools"
+    if not tools and not tool_schemas:
+        tools = True
+
+    # Possibly interpret the instruction to refine it
+    instruction_str = None
+    if interpret:
+        instruction_str = await branch.interpret(
+            str(
+                instruct.to_dict()
+                if isinstance(instruct, Instruct)
+                else instruct
+            )
+        )
+
+    # Convert Instruct to dict if necessary
+    instruct_dict = (
+        instruct.to_dict()
+        if isinstance(instruct, Instruct)
+        else dict(instruct)
+    )
+    # Overwrite the "instruction" field with the interpreted string (if any)
+    instruct_dict["instruction"] = instruction_str or instruct_dict.get(
+        "instruction"
+    )
+
+    # Prepare a copy of user-provided kwargs for the first operate call
+    kwargs_for_operate = copy(kwargs)
+    kwargs_for_operate["actions"] = True
+    kwargs_for_operate["reason"] = True
+
+    # We'll pass the refined instruct_dict plus the user’s other kwargs
+    from .utils import ReActAnalysis
+
+    # Step 1: Generate initial ReAct analysis
+    analysis: ReActAnalysis = await branch.operate(
+        response_format=ReActAnalysis,
+        tools=tools,
+        tool_schemas=tool_schemas,
+        chat_model=analysis_model or branch.chat_model,
+        **kwargs_for_operate,
+    )
+    analyses = [analysis]
+
+    # Validate and clamp max_extensions if needed
+    if max_extensions and max_extensions > 5:
+        logging.warning("max_extensions should not exceed 5; defaulting to 5.")
+        max_extensions = 5
+
+    # Step 2: Possibly loop through expansions if extension_needed
+    extensions = max_extensions
+    while (
+        extension_allowed
+        and analysis.extension_needed
+        and (extensions if extensions else 1) > 0
+    ):
+        new_instruction = None
+        if extensions == max_extensions:
+            new_instruction = ReActAnalysis.FIRST_EXT_PROMPT.format(
+                extensions=extensions
+            )
+        else:
+            new_instruction = ReActAnalysis.CONTINUE_EXT_PROMPT.format(
+                extensions=extensions
+            )
+
+        # Each expansion uses a fresh copy of instruct_dict + forcibly "reason" + "actions"
+        expanded_kwargs = copy(instruct_dict)
+        expanded_kwargs["instruction"] = new_instruction
+        expanded_kwargs["reason"] = True
+        expanded_kwargs["actions"] = True
+
+        analysis = await branch.operate(
+            response_format=ReActAnalysis,
+            tools=tools,
+            tool_schemas=tool_schemas,
+            **expanded_kwargs,
+        )
+        analyses.append(analysis)
+
+        if extensions:
+            extensions -= 1
+
+    # Step 3: Produce final answer by calling branch._instruct with an answer prompt
+    answer_prompt = ReActAnalysis.ANSWER_PROMPT.format(
+        instruction=instruct_dict["instruction"]
+    )
+    out = await branch.communicate(
+        instruction=answer_prompt,
+        response_format=response_format,
+        **(response_kwargs or {}),
+    )
+    if return_analysis:
+        return out, analyses
+    return out

lionagi/operations/ReAct/utils.py

@@ -0,0 +1,28 @@
+# Copyright (c) 2023 - 2024, HaiyangLi <quantocean.li at gmail dot com>
+#
+# SPDX-License-Identifier: Apache-2.0
+
+from typing import ClassVar
+
+from pydantic import BaseModel, Field
+
+
+class ReActAnalysis(BaseModel):
+
+    FIRST_EXT_PROMPT: ClassVar[str] = (
+        "You are provided with another round to perform reason action to provide an accurate final answer. you have max another {extensions} rounds, set extension_needed to False if you are done and ready to provide final answer."
+    )
+
+    CONTINUE_EXT_PROMPT: ClassVar[str] = (
+        "You are provided with another round, you have max another {extensions} rounds"
+    )
+
+    ANSWER_PROMPT: ClassVar[str] = (
+        "given above reason and actions, please provide final answer to the original user request {instruction}"
+    )
+
+    analysis: str
+    extension_needed: bool = Field(
+        False,
+        description="Set to True if more steps are needed to provide an accurate answer. If True, additional rounds are allowed.",
+    )

lionagi/operations/__init__.py

@@ -2,12 +2,4 @@
 #
 # SPDX-License-Identifier: Apache-2.0
 
-from .brainstorm.brainstorm import brainstorm
-from .plan.plan import plan
-from .select.select import select
-
-__all__ = (
-    "brainstorm",
-    "plan",
-    "select",
-)
+from .types import *

lionagi/operations/_act/act.py

@@ -0,0 +1,73 @@
+# Copyright (c) 2023 - 2024, HaiyangLi <quantocean.li at gmail dot com>
+#
+# SPDX-License-Identifier: Apache-2.0
+
+import logging
+from typing import TYPE_CHECKING
+
+from pydantic import BaseModel
+
+from lionagi.protocols.types import ActionResponse, Log
+
+if TYPE_CHECKING:
+    from lionagi.session.branch import Branch
+
+
+async def _act(
+    branch: "Branch",
+    action_request: BaseModel | dict,
+    suppress_errors: bool = False,
+) -> ActionResponse:
+
+    _request = {}
+
+    if isinstance(action_request, BaseModel):
+        if hasattr(action_request, "function") and hasattr(
+            action_request, "arguments"
+        ):
+            _request["function"] = action_request.function
+            _request["arguments"] = action_request.arguments
+    elif isinstance(action_request, dict):
+        if {"function", "arguments"} <= set(action_request.keys()):
+            _request["function"] = action_request["function"]
+            _request["arguments"] = action_request["arguments"]
+
+    try:
+        func_call = await branch._action_manager.invoke(_request)
+    except Exception as e:
+        branch._log_manager.log(Log(content={"error": str(e)}))
+        if suppress_errors:
+            logging.error(
+                f"Error invoking action '{_request['function']}': {e}"
+            )
+            return None
+        raise e
+
+    branch._log_manager.log(Log.create(func_call))
+
+    from lionagi.protocols.types import ActionRequest
+
+    if not isinstance(action_request, ActionRequest):
+        action_request = ActionRequest.create(
+            sender=branch.id,
+            recipient=func_call.func_tool.id,
+            **_request,
+        )
+
+    # Add the action request/response to the message manager, if not present
+    if action_request not in branch.messages:
+        branch.msgs.add_message(action_request=action_request)
+
+    branch.msgs.add_message(
+        action_request=action_request,
+        action_output=func_call.response,
+    )
+
+    # Return an ActionResponse object
+    from lionagi.operatives.types import ActionResponseModel
+
+    return ActionResponseModel(
+        function=action_request.function,
+        arguments=action_request.arguments,
+        output=func_call.response,
+    )

lionagi/operations/chat/__init__.py

@@ -0,0 +1,3 @@
+# Copyright (c) 2023 - 2024, HaiyangLi <quantocean.li at gmail dot com>
+#
+# SPDX-License-Identifier: Apache-2.0