flock-core 0.4.527__py3-none-any.whl → 0.5.0b0__py3-none-any.whl

flock/modules/mem0_async/async_mem0_module.py

@@ -1,126 +0,0 @@
-from typing import Any
-
-# from mem0.client.main import AsyncMemoryClient, MemoryClient
-# from mem0.memory.main import AsyncMemory
-from mem0 import AsyncMemory, AsyncMemoryClient
-from pydantic import Field
-
-from flock.core.context.context import FlockContext
-from flock.core.flock_agent import FlockAgent
-from flock.core.flock_module import FlockModule, FlockModuleConfig
-from flock.core.flock_registry import flock_component
-from flock.core.logging.logging import get_logger
-
-logger = get_logger("module.mem0")
-
-
-config = {
-    "vector_store": {
-        "provider": "chroma",
-        "config": {
-            "collection_name": "flock_memory",
-            "path": ".flock/memory",
-        }
-    }
-}
-
-
-class AsyncMem0ModuleConfig(FlockModuleConfig):
-    top_k: int = Field(default=10, description="Number of memories to retrieve")
-    user_id: str = Field(default="flock", description="User ID the memories will be associated with")
-    agent_id: str = Field(default="flock", description="Agent ID the memories will be associated with")
-    memory_input_key: str | None = Field(default=None, description="Input key to use for memory, if none the description of the agent will be used")
-    api_key: str | None = Field(default=None, description="API key for mem0 Platform")
-    config: dict[str, Any] = Field(default=config, description="Configuration for mem0")
-
-
-@flock_component(config_class=AsyncMem0ModuleConfig)
-class AsyncMem0Module(FlockModule):
-
-    name: str = "mem0"
-    config: AsyncMem0ModuleConfig = AsyncMem0ModuleConfig()
-
-
-    def __init__(self, name, config: AsyncMem0ModuleConfig) -> None:
-        global memory
-        """Initialize Mem0 module."""
-        super().__init__(name=name, config=config)
-        logger.debug("Initializing Mem0 module")
-
-
-
-
-    def dict_to_str_repr(self,d: dict) -> str:
-        return repr(d)
-
-
-    async def on_post_evaluate(
-        self,
-        agent: FlockAgent,
-        inputs: dict[str, Any],
-        context: FlockContext | None = None,
-        result: dict[str, Any] | None = None,
-    ) -> dict[str, Any]:
-        if self.config.api_key:
-            memory = AsyncMemoryClient(api_key=self.config.api_key)
-        else:
-            memory = await AsyncMemory.from_config(config_dict=self.config.config)
-
-        agent_id = self.config.agent_id if self.config.agent_id else agent.name
-
-        # get the result without the inputs
-        filtered_result = {k: v for k, v in result.items() if k not in inputs}
-        # get the inputs without memory
-        filtered_inputs = {k: v for k, v in inputs.items() if k not in [self.config.memory_input_key]}
-
-        # add memories about the user inputs
-        added_user_memory = await memory.add(self.dict_to_str_repr(filtered_inputs), user_id=self.config.user_id)
-        logger.info(f"Added caller memory: {added_user_memory}")
-
-        # add memories about the agent result
-        added_agent_memory = await memory.add(self.dict_to_str_repr(filtered_result), agent_id=agent_id)
-        logger.info(f"Added agent memory: {added_agent_memory}")
-
-
-        return result
-
-    async def on_pre_evaluate(
-        self,
-        agent: FlockAgent,
-        inputs: dict[str, Any],
-        context: FlockContext | None = None,
-    ) -> dict[str, Any]:
-        if self.config.api_key:
-            memory = AsyncMemoryClient(api_key=self.config.api_key)
-        else:
-            memory = await AsyncMemory.from_config(config_dict=self.config.config)
-
-        message = self.dict_to_str_repr(inputs)
-        agent_id = self.config.agent_id if self.config.agent_id else agent.name
-
-        relevant_agent_memories = await memory.search(query=message, agent_id=agent_id, limit=self.config.top_k)
-        logger.info(f"Relevant agent memories: {relevant_agent_memories}")
-
-        relevant_user_memories = await memory.search(query=message, user_id=self.config.user_id, limit=self.config.top_k)
-        logger.info(f"Relevant user memories: {relevant_user_memories}")
-
-        if relevant_agent_memories or relevant_user_memories:
-            memories_str = ''
-            if "results" in relevant_agent_memories:
-                memories_str = "\n".join(f"- {entry['memory']}" for entry in relevant_agent_memories["results"])
-            else:
-                memories_str = "\n".join(f"- {entry}" for entry in relevant_agent_memories)
-
-            if "results" in relevant_user_memories:
-                memories_str = memories_str + "\n" + "\n".join(f"- {entry['memory']}" for entry in relevant_user_memories["results"])
-            else:
-                memories_str = memories_str + "\n" + "\n".join(f"- {entry}" for entry in relevant_user_memories)
-
-            if memories_str:
-                if self.config.memory_input_key:
-                    inputs[self.config.memory_input_key] = memories_str
-                else:
-                    agent.description = agent.description + "\n\n Memories:" + memories_str
-
-
-        return inputs

flock/modules/memory/__init__.py

@@ -1 +0,0 @@
-# Package for modules

flock/modules/memory/memory_module.py

@@ -1,429 +0,0 @@
-import json
-import uuid
-from datetime import datetime
-from typing import Any, Literal
-
-from pydantic import Field
-
-from flock.core.context.context import FlockContext
-
-# if TYPE_CHECKING:
-#     from flock.core import FlockAgent
-from flock.core.flock_agent import FlockAgent
-from flock.core.flock_module import FlockModule, FlockModuleConfig
-from flock.core.flock_registry import flock_component
-from flock.core.logging.logging import get_logger
-from flock.modules.memory.memory_parser import MemoryMappingParser
-from flock.modules.memory.memory_storage import FlockMemoryStore, MemoryEntry
-
-logger = get_logger("memory")
-
-
-class MemoryModuleConfig(FlockModuleConfig):
-    """Configuration for the MemoryModule.
-
-    This class defines the configuration for the MemoryModule.
-    """
-
-    folder_path: str = Field(
-        default=".flock/memory/",
-        description="Directory where memory file and concept graph will be saved",
-    )
-    concept_graph_file: str = Field(
-        default="concept_graph.png",
-        description="Base filename for the concept graph image",
-    )
-    file_path: str | None = Field(
-        default="agent_memory.json", description="Path to save memory file"
-    )
-    memory_mapping: str | None = Field(
-        default=None, description="Memory mapping configuration"
-    )
-    similarity_threshold: float = Field(
-        default=0.5, description="Threshold for semantic similarity"
-    )
-    max_length: int = Field(
-        default=1000, description="Max length of memory entry before splitting"
-    )
-    save_after_update: bool = Field(
-        default=True, description="Whether to save memory after each update"
-    )
-    splitting_mode: Literal["summary", "semantic", "characters", "none"] = (
-        Field(default="characters", description="Mode to split memory content")
-    )
-    enable_read_only_mode: bool = Field(
-        default=False, description="Whether to enable read only mode"
-    )
-    enable_write_only_mode: bool = Field(
-        default=False, description="Whether to enable write only mode"
-    )
-    number_of_concepts_to_extract: int = Field(
-        default=3, description="Number of concepts to extract from the memory"
-    )
-    memory_input_key: str | None = Field(default=None, description="Input key to use for memory, if none the description of the agent will be used")
-
-
-
-@flock_component(config_class=MemoryModuleConfig)
-class MemoryModule(FlockModule):
-    """Module that adds memory capabilities to a Flock agent."""
-
-    name: str = "memory"
-    config: MemoryModuleConfig = Field(
-        default_factory=MemoryModuleConfig,
-        description="Memory module configuration",
-    )
-    memory_store: FlockMemoryStore | None = None
-    memory_ops: list[Any] = []
-
-    def __init__(self, name: str, config: MemoryModuleConfig):
-        super().__init__(name=name, config=config)
-        self.memory_store = FlockMemoryStore.load_from_file(
-            self.get_memory_filename(name)
-        )
-        self.memory_ops = (
-            MemoryMappingParser().parse(self.config.memory_mapping)
-            if self.config.memory_mapping
-            else [{"type": "semantic"}]
-        )
-
-    async def on_initialize(
-        self,
-        agent: FlockAgent,
-        inputs: dict[str, Any],
-        context: FlockContext | None = None,
-    ) -> None:
-        """Initialize memory store if needed."""
-        if not self.memory_store:
-            self.memory_store = FlockMemoryStore.load_from_file(
-                self.get_memory_filename(self.name)
-            )
-        self.memory_ops = (
-            MemoryMappingParser().parse(self.config.memory_mapping)
-            if self.config.memory_mapping
-            else [{"type": "semantic"}]
-        )
-        logger.debug(f"Initialized memory module for agent {agent.name}")
-
-    async def on_pre_evaluate(
-        self,
-        agent: FlockAgent,
-        inputs: dict[str, Any],
-        context: FlockContext | None = None,
-    ) -> dict[str, Any]:
-        """Check memory before evaluation."""
-        if not self.memory_store:
-            return inputs
-
-        if self.config.enable_write_only_mode:
-            return inputs
-
-        inputs = await self.search_memory(agent, inputs)
-
-        if "context" in inputs:
-            agent.input = (
-                agent.input + ", context: list | context with more information"
-            )
-
-        return inputs
-
-    def get_memory_filename(self, module_name: str) -> str:
-        """Generate the full file path for the memory file."""
-        folder = self.config.folder_path
-        if not folder.endswith(("/", "\\")):
-            folder += "/"
-        import os
-
-        if not os.path.exists(folder):
-            os.makedirs(folder, exist_ok=True)
-        # Determine base filename and extension from file_path config
-        if self.config.file_path:
-            file_name = self.config.file_path.rsplit("/", 1)[-1].rsplit(
-                "\\", 1
-            )[-1]
-            if "." in file_name:
-                base, ext = file_name.rsplit(".", 1)
-                ext = f".{ext}"
-            else:
-                base, ext = file_name, ""
-        else:
-            base, ext = "agent_memory", ".json"
-        return f"{folder}{module_name}_{base}{ext}"
-
-    def get_concept_graph_filename(self, module_name: str) -> str:
-        """Generate the full file path for the concept graph image."""
-        folder = self.config.folder_path
-        if not folder.endswith(("/", "\\")):
-            folder += "/"
-        import os
-
-        if not os.path.exists(folder):
-            os.makedirs(folder, exist_ok=True)
-        # Use timestamp to create a unique filename
-        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S_%f")[:-3]
-        if self.config.concept_graph_file:
-            file_name = self.config.concept_graph_file.rsplit("/", 1)[
-                -1
-            ].rsplit("\\", 1)[-1]
-            if "." in file_name:
-                base, ext = file_name.rsplit(".", 1)
-                ext = f".{ext}"
-            else:
-                base, ext = file_name, ""
-        else:
-            base, ext = "concept_graph", ".png"
-        return f"{folder}{module_name}_{base}_{timestamp}{ext}"
-
-    async def search_memory(
-        self, agent: FlockAgent, query: dict[str, Any]
-    ) -> dict[str, Any]:
-        """Search memory for the query."""
-        if not self.memory_store:
-            # No memory store loaded – just return the untouched input
-            return query
-
-        try:
-            input_text = json.dumps(query)
-            query_embedding = self.memory_store.compute_embedding(input_text)
-            concepts = await self._extract_concepts(
-                agent, input_text, self.config.number_of_concepts_to_extract
-            )
-
-            memory_results = []
-            for op in self.memory_ops:
-                if op["type"] == "semantic":
-                    semantic_results = self.memory_store.retrieve(
-                        query_embedding,
-                        concepts,
-                        similarity_threshold=self.config.similarity_threshold,
-                    )
-                    memory_results.extend(semantic_results)
-                elif op["type"] == "exact":
-                    exact_results = self.memory_store.exact_match(query)
-                    memory_results.extend(exact_results)
-
-            context: list[dict[str, Any]] = []
-            if memory_results:
-                for result in memory_results:
-                    context.append(
-                        {"content": result.content, "concepts": result.concepts}
-                    )
-
-                logger.debug(
-                    f"Found {len(memory_results)} relevant memories",
-                    agent=agent.name,
-                )
-                query["context"] = context
-
-            return query
-
-        except Exception as e:
-            logger.warning(f"Memory retrieval failed: {e}", agent=agent.name)
-            return query
-
-    async def add_to_memory(
-        self, agent: FlockAgent, data: dict[str, Any]
-    ) -> None:
-        """Add data to memory."""
-        if not self.memory_store:
-            return
-
-        try:
-            chunks = await self._get_chunks(agent, data, None)
-            await self._store_chunks(agent, chunks)
-        except Exception as e:
-            logger.warning(f"Memory storage failed: {e}", agent=agent.name)
-
-    async def on_post_evaluate(
-        self,
-        agent: FlockAgent,
-        inputs: dict[str, Any],
-        context: FlockContext | None = None,
-        result: dict[str, Any] | None = None,
-    ) -> dict[str, Any]:
-        """Store results in memory after evaluation."""
-        if not self.memory_store:
-            return result
-
-        if self.config.enable_read_only_mode:
-            return result
-
-        try:
-            chunks = await self._get_chunks(agent, inputs, result)
-            await self._store_chunks(agent, chunks)
-        except Exception as e:
-            logger.warning(f"Memory storage failed: {e}", agent=agent.name)
-
-        return result
-
-    async def on_terminate(
-        self,
-        agent: Any,
-        inputs: dict[str, Any],
-        result: dict[str, Any],
-        context: FlockContext | None = None,
-    ) -> None:
-        """Save memory store if configured."""
-        if self.config.save_after_update and self.memory_store:
-            self.save_memory()
-
-    async def _extract_concepts(
-        self, agent: FlockAgent, text: str, number_of_concepts: int = 3
-    ) -> set[str]:
-        """Extract concepts using the agent's LLM capabilities."""
-        existing_concepts = set()
-        if self.memory_store and self.memory_store.concept_graph:
-            existing_concepts = set(
-                self.memory_store.concept_graph.graph.nodes()
-            )
-
-        input_signature = "text: str | Text to analyze"
-        if existing_concepts:
-            input_signature += ", existing_concepts: list[str] | Already known concepts that might apply"
-
-        concept_signature = agent.create_dspy_signature_class(
-            f"{agent.name}_concept_extractor",
-            "Extract key concepts from text",
-            f"{input_signature} -> concepts: list[str] | Max {number_of_concepts} key concepts all lower case",
-        )
-
-        agent._configure_language_model(agent.model, True, 0.0, 8192)
-        predictor = agent._select_task(concept_signature, "Completion")
-        result_obj = predictor(
-            text=text,
-            existing_concepts=list(existing_concepts)
-            if existing_concepts
-            else None,
-        )
-        concept_list = getattr(result_obj, "concepts", [])
-        return set(concept_list)
-
-    async def _summarize_mode(
-        self,
-        agent: FlockAgent,
-        inputs: dict[str, Any],
-        result: dict[str, Any],
-    ) -> str:
-        """Extract information chunks using summary mode."""
-        split_signature = agent.create_dspy_signature_class(
-            f"{agent.name}_splitter",
-            "Extract a list of potentially needed data and information for future reference",
-            """
-            content: str | The content to split
-            -> chunks: list[str] | List of data and information for future reference
-            """,
-        )
-        agent._configure_language_model(agent.model, True, 0.0, 8192)
-        splitter = agent._select_task(split_signature, "Completion")
-        full_text = json.dumps(inputs) + json.dumps(result)
-        split_result = splitter(content=full_text)
-        return "\n".join(split_result.chunks)
-
-    async def _semantic_splitter_mode(
-        self,
-        agent: FlockAgent,
-        inputs: dict[str, Any],
-        result: dict[str, Any],
-    ) -> list[str]:
-        """Extract information chunks using semantic mode."""
-        split_signature = agent.create_dspy_signature_class(
-            f"{self.name}_splitter",
-            "Split content into meaningful, self-contained chunks",
-            """
-            content: str | The content to split
-            -> chunks: list[dict[str,str]] | List of chunks as key-value pairs - keys are a short title and values are the chunk content
-            """,
-        )
-        agent._configure_language_model(agent.model, True, 0.0, 8192)
-        splitter = agent._select_task(split_signature, "Completion")
-        full_text = json.dumps(inputs) + (json.dumps(result) if result else "")
-        split_result = splitter(content=full_text)
-        # Flatten list[dict] into list[str] of "title: content" strings to
-        # keep downstream storage logic simple and type-safe.
-        flattened: list[str] = []
-        for chunk in split_result.chunks:
-            if isinstance(chunk, dict):
-                flattened.extend([f"{k}: {v}" for k, v in chunk.items()])
-            else:
-                flattened.append(str(chunk))
-        return flattened
-
-    async def _character_splitter_mode(
-        self,
-        agent: FlockAgent,
-        inputs: dict[str, Any],
-        result: dict[str, Any],
-    ) -> list[str]:
-        """Extract information chunks by splitting text into fixed character lengths."""
-        full_text = json.dumps(inputs) + (json.dumps(result) if result else "")
-        return [
-            full_text[i : i + self.config.max_length]
-            for i in range(0, len(full_text), self.config.max_length)
-        ]
-
-    async def _get_chunks(
-        self,
-        agent: FlockAgent,
-        inputs: dict[str, Any],
-        result: dict[str, Any] | None,
-    ) -> str | list[str]:
-        """Get memory chunks based on the configured splitting mode."""
-        mode = self.config.splitting_mode
-        if mode == "semantic":
-            return await self._semantic_splitter_mode(agent, inputs, result)
-        elif mode == "summary":
-            return await self._summarize_mode(agent, inputs, result)
-        elif mode == "characters":
-            return await self._character_splitter_mode(agent, inputs, result)
-        elif mode == "none":
-            return (
-                json.dumps(inputs) + json.dumps(result)
-                if result
-                else json.dumps(inputs)
-            )
-        else:
-            raise ValueError(f"Unknown splitting mode: {mode}")
-
-    async def _store_chunk(self, agent: FlockAgent, chunk: str) -> None:
-        """Store a single chunk in memory."""
-        chunk_concepts = await self._extract_concepts(
-            agent, chunk, self.config.number_of_concepts_to_extract
-        )
-        entry = MemoryEntry(
-            id=str(uuid.uuid4()),
-            content=chunk,
-            embedding=self.memory_store.compute_embedding(chunk).tolist(),
-            concepts=chunk_concepts,
-            timestamp=datetime.now(),
-        )
-        self.memory_store.add_entry(entry)
-        if self.config.save_after_update:
-            self.save_memory()
-        logger.debug(
-            "Stored interaction in memory",
-            agent=agent.name,
-            entry_id=entry.id,
-            concepts=chunk_concepts,
-        )
-
-    async def _store_chunks(
-        self, agent: FlockAgent, chunks: str | list[str]
-    ) -> None:
-        """Store chunks (single or multiple) in memory."""
-        if isinstance(chunks, str):
-            await self._store_chunk(agent, chunks)
-        elif isinstance(chunks, list):
-            # Avoid tqdm in async context – simple for-loop is safer.
-            for chunk in chunks:
-                await self._store_chunk(agent, chunk)
-
-    def save_memory(self) -> None:
-        """Save memory store to file."""
-        if self.memory_store and self.config.file_path:
-            json_str = self.memory_store.model_dump_json()
-            filename = self.get_memory_filename(self.name)
-            with open(filename, "w") as file:
-                file.write(json_str)
-            self.memory_store.concept_graph.save_as_image(
-                self.get_concept_graph_filename(self.name)
-            )

flock/modules/memory/memory_parser.py

@@ -1,125 +0,0 @@
-"""Parser for memory mapping declarations into executable operations."""
-
-import re
-from typing import Any
-
-from flock.modules.memory.memory_storage import (
-    CombineOperation,
-    EnrichOperation,
-    ExactOperation,
-    FilterOperation,
-    MemoryOperation,
-    MemoryScope,
-    SemanticOperation,
-    SortOperation,
-)
-
-
-class MemoryMappingParser:
-    """Parses memory mapping declarations into executable operations."""
-
-    def parse(self, mapping: str) -> list[MemoryOperation]:
-        """Parse a memory mapping string into operations.
-
-        Example mappings:
-        "topic -> memory.semantic(threshold=0.9) | memory.exact -> output"
-        "query -> memory.semantic(scope='global') | memory.filter(recency='7d') | memory.sort(by='relevance')"
-        """
-        operations = []
-        stages = [s.strip() for s in mapping.split("|")]
-
-        for stage in stages:
-            if "->" not in stage:
-                continue
-
-            inputs, op_spec = stage.split("->")
-            inputs = [i.strip() for i in inputs.split(",")]
-
-            if "memory." in op_spec:
-                # Extract operation name and parameters
-                match = re.match(r"memory\.(\w+)(?:\((.*)\))?", op_spec.strip())
-                if not match:
-                    continue
-
-                op_name, params_str = match.groups()
-                params = self._parse_params(params_str or "")
-
-                # Create appropriate operation object
-                if op_name == "semantic":
-                    operation = SemanticOperation(
-                        threshold=params.get("threshold", 0.8),
-                        scope=params.get("scope", MemoryScope.BOTH),
-                        max_results=params.get("max_results", 10),
-                    )
-                elif op_name == "exact":
-                    operation = ExactOperation(
-                        keys=inputs, scope=params.get("scope", MemoryScope.BOTH)
-                    )
-                elif op_name == "enrich":
-                    operation = EnrichOperation(
-                        tools=params.get("tools", []),
-                        strategy=params.get("strategy", "comprehensive"),
-                        scope=params.get("scope", MemoryScope.BOTH),
-                    )
-                elif op_name == "filter":
-                    operation = FilterOperation(
-                        recency=params.get("recency"),
-                        relevance=params.get("relevance"),
-                        metadata=params.get("metadata", {}),
-                    )
-                elif op_name == "sort":
-                    operation = SortOperation(
-                        by=params.get("by", "relevance"),
-                        ascending=params.get("ascending", False),
-                    )
-                elif op_name == "combine":
-                    operation = CombineOperation(
-                        weights=params.get(
-                            "weights", {"semantic": 0.7, "exact": 0.3}
-                        )
-                    )
-
-                operations.append(operation)
-
-        return operations
-
-    def _parse_params(self, params_str: str) -> dict[str, Any]:
-        """Parse parameters string into a dictionary.
-
-        Handles:
-        - Quoted strings: threshold='high'
-        - Numbers: threshold=0.9
-        - Lists: tools=['web_search', 'extract_numbers']
-        - Dictionaries: weights={'semantic': 0.7, 'exact': 0.3}
-        """
-        if not params_str:
-            return {}
-
-        params = {}
-        # Split on commas not inside brackets or quotes
-        param_pairs = re.findall(
-            r"""
-            (?:[^,"]|"[^"]*"|'[^']*')+  # Match everything except comma, or quoted strings
-        """,
-            params_str,
-            re.VERBOSE,
-        )
-
-        for pair in param_pairs:
-            if "=" not in pair:
-                continue
-            key, value = pair.split("=", 1)
-            key = key.strip()
-            value = value.strip()
-
-            # Try to evaluate the value (for lists, dicts, numbers)
-            try:
-                # Safely evaluate the value
-                value = eval(value, {"__builtins__": {}}, {})
-            except:
-                # If evaluation fails, treat as string
-                value = value.strip("'\"")
-
-            params[key] = value
-
-        return params