Interface defining the global configuration settings and behaviors for the swarm system.
Centralizes constants and functions used across components like ClientAgent (e.g., tool handling, logging, history).
Influences workflows such as message processing (CC_EMPTY_OUTPUT_PLACEHOLDERS in RUN_FN), tool call recovery (CC_RESQUE_STRATEGY in _resurrectModel), and logging (CC_LOGGER_ENABLE_DEBUG).
Customizable via setConfig to adapt system behavior dynamically.
CC_TOOL_CALL_EXCEPTION_FLUSH_PROMPT: string
A prompt used to flush the conversation when tool call exceptions occur, specifically for troubleshooting in llama3.1:8b models.
Applied in ClientAgent._resurrectModel with the "flush" strategy to reset the conversation state. Requires CC_OLLAMA_EMIT_TOOL_PROTOCOL to be disabled.
CC_TOOL_CALL_EXCEPTION_RECOMPLETE_PROMPT: string
A multi-line prompt to recomplete invalid tool calls, designed as a fix for intermittent issues in IlyaGusev/saiga_yandexgpt_8b_gguf (LMStudio).
Used in ClientAgent.getCompletion with the "recomplete" strategy, instructing the model to analyze, correct, and explain tool call errors.
CC_EMPTY_OUTPUT_PLACEHOLDERS: string[]
An array of placeholder responses for empty model outputs, used in ClientAgent.createPlaceholder to greet or prompt the user.
Randomly selected in ClientAgent._resurrectModel or RUN_FN when output is empty, enhancing user experience by avoiding silent failures.
CC_KEEP_MESSAGES: number
Maximum number of messages to retain in history, used indirectly in ClientAgent.history management.
Limits history to 15 messages, though not explicitly enforced in provided ClientAgent code.
CC_MAX_TOOL_CALLS: number
Maximum number of tool calls allowed per execution, used in ClientAgent.EXECUTE_FN to cap toolCalls.
Limits to 1 tool call by default, preventing excessive tool invocation loops in a single run.
CC_AGENT_MAP_TOOLS: (tool: IToolCall[], clientId: string, agentName: string) => IToolCall[] | Promise<IToolCall[]>
Function to map tool calls for an agent, used in ClientAgent.mapToolCalls (e.g., EXECUTE_FN).
Default implementation returns tools unchanged, allowing customization of IToolCall processing via setConfig.
CC_GET_AGENT_HISTORY_ADAPTER: (clientId: string, agentName: string) => IHistoryAdapter
Factory function to provide a history adapter for an agent, used in ClientAgent.history (e.g., getCompletion).
Returns HistoryAdapter by default, implementing IHistoryAdapter for message storage and retrieval.
CC_GET_CLIENT_LOGGER_ADAPTER: () => ILoggerAdapter
Factory function to provide a logger adapter for clients, used in ClientAgent.logger (e.g., debug logging).
Returns LoggerAdapter by default, implementing ILoggerAdapter for logging control across the system.
CC_SWARM_AGENT_CHANGED: (clientId: string, agentName: string, swarmName: string) => Promise<void>
Callback function triggered when the active agent changes in a swarm, used in swarm-related logic (e.g., ISwarmParams).
Default implementation is a no-op, observed indirectly in ClientAgent.commitAgentChange via IBus.emit "commit-agent-change".
CC_SWARM_DEFAULT_AGENT: (clientId: string, swarmName: string, defaultAgent: string) => Promise<string>
Function to determine the default agent for a swarm, used in swarm initialization (e.g., ISwarmParams).
Default implementation returns the provided defaultAgent, aligning with swarm-agent hierarchy logic, though not directly observed in ClientAgent.
CC_SWARM_DEFAULT_STACK: (clientId: string, swarmName: string) => Promise<string[]>
Function to provide the default navigation stack for a swarm, used in ISwarmParams initialization.
Default implementation returns an empty array, part of swarm navigation setup, though not directly used in ClientAgent.
CC_SWARM_STACK_CHANGED: (clientId: string, navigationStack: string[], swarmName: string) => Promise<void>
Callback function triggered when the navigation stack changes in a swarm, used in ISwarmParams (e.g., navigationPop).
Default implementation is a no-op, indirectly related to ClientAgent.commitAgentChange for stack updates.
CC_AGENT_DEFAULT_VALIDATION: (output: string) => string | Promise<string>
Default validation function for agent outputs, used in ClientAgent.validate (e.g., RUN_FN, EXECUTE_FN).
Imported from validateDefault, returns null if valid or an error string if invalid, ensuring output correctness.
CC_AGENT_HISTORY_FILTER: (agentName: string) => (message: IModelMessage<object>) => boolean
Filter function for agent history, used in ClientAgent.history.toArrayForAgent to scope messages.
Ensures only relevant messages (e.g., matching agentName for "tool" or tool_calls) are included in completions.
CC_AGENT_OUTPUT_TRANSFORM: (input: string, clientId: string, agentName: string) => string | Promise<string>
Default transformation function for agent outputs, used in ClientAgent.transform (e.g., RUN_FN, _emitOutput).
Removes XML tags via removeXmlTags based on CC_AGENT_DISALLOWED_TAGS to clean responses for consistency.
CC_AGENT_OUTPUT_MAP: (message: IModelMessage<object>) => IModelMessage<object> | Promise<IModelMessage<object>>
Function to map model messages for agent output, used in ClientAgent.map (e.g., RUN_FN, EXECUTE_FN).
Default implementation returns the message unchanged, allowing customization of IModelMessage via setConfig.
CC_AGENT_SYSTEM_PROMPT: string[]
Optional system prompt for agents, used in ClientAgent.history.toArrayForAgent (e.g., getCompletion).
Undefined by default, allowing optional agent-specific instructions to be added to history via setConfig.
CC_AGENT_DISALLOWED_TAGS: string[]
Array of XML tags disallowed in agent outputs, used with CC_AGENT_OUTPUT_TRANSFORM in ClientAgent.transform.
Filters out tags like "tool_call" via removeXmlTags in RUN_FN to clean responses for downstream processing.
CC_AGENT_DISALLOWED_SYMBOLS: string[]
Array of symbols disallowed in agent outputs, potentially used in validation or transformation logic.
Includes curly braces, suggesting filtering of JSON-like structures, though not directly observed in ClientAgent.
CC_STORAGE_SEARCH_SIMILARITY: number
Similarity threshold for storage searches, used in IStorage.take for similarity-based retrieval.
Set to 0.65, defining the minimum similarity score for search results, though not directly in ClientAgent.
CC_STORAGE_SEARCH_POOL: number
Maximum number of results for storage searches, used in IStorage.take to limit retrieval.
Caps search pool at 5 items by default, though not directly observed in ClientAgent.
CC_LOGGER_ENABLE_INFO: boolean
Flag to enable info-level logging, used in ClientAgent.logger for informational logs.
Disabled by default (false), controlling verbosity of ILoggerAdapter logs.
CC_LOGGER_ENABLE_DEBUG: boolean
Flag to enable debug-level logging, used extensively in ClientAgent.logger.debug (e.g., RUN_FN, EXECUTE_FN).
Disabled by default (false), gating detailed debug output in ILoggerAdapter.
CC_LOGGER_ENABLE_LOG: boolean
Flag to enable general logging, used in ClientAgent.logger for basic log output.
Enabled by default (true), ensuring core logging functionality in ILoggerAdapter.
CC_LOGGER_ENABLE_CONSOLE: boolean
Flag to enable console logging, used in ClientAgent.logger for direct console output.
Disabled by default (false), allowing logs to be redirected via ILoggerAdapter.
CC_RESQUE_STRATEGY: "flush" | "recomplete" | "custom"
Strategy for handling model resurrection, used in ClientAgent._resurrectModel and getCompletion.
Options: "flush" (reset conversation), "recomplete" (retry tool calls), "custom" (user-defined); determines recovery approach for invalid outputs or tool calls.
CC_NAME_TO_TITLE: (name: string) => string
Utility function to convert names to title case, used for UI or logging readability.
Imported from nameToTitle, enhancing presentation of agent or swarm names, though not directly in ClientAgent.
CC_FN_PLANTUML: (uml: string) => Promise<string>
Function to process PlantUML diagrams, potentially for visualization purposes.
Default returns an empty string, allowing custom UML rendering logic via setConfig, though not directly in ClientAgent.
CC_PROCESS_UUID: string
Unique identifier for the current process, used system-wide for tracking or logging.
Generated via randomString, providing a process-specific UUID, though not directly in ClientAgent.
CC_BANHAMMER_PLACEHOLDER: string
Placeholder response for banned topics or actions, used in IPolicy.banClient enforcement.
Indicates refusal to engage, enhancing policy messaging, though not directly in ClientAgent.
CC_TOOL_CALL_EXCEPTION_CUSTOM_FUNCTION: (clientId: string, agentName: string) => Promise<IModelMessage<object>>
A custom function to handle tool call exceptions by returning a model message or null, used in ClientAgent.getCompletion with the "custom" CC_RESQUE_STRATEGY.
Default implementation returns null, allowing override via setConfig to implement specific recovery logic tailored to the application.
CC_PERSIST_ENABLED_BY_DEFAULT: boolean
Flag to enable persistence by default, used in IStorage or IState initialization.
Enabled (true) by default, suggesting data retention unless overridden, though not directly in ClientAgent.
CC_AUTOBAN_ENABLED_BY_DEFAULT: boolean
Flag to enable autobanning by default, used in IPolicy for automatic ban enforcement.
Disabled (false) by default, allowing manual ban control unless overridden, though not directly in ClientAgent.
CC_DEFAULT_STATE_SET: <T = any>(state: T, clientId: string, stateName: string) => Promise<void>
Default function to set state values, used in IState.setState for state persistence.
No-op by default, allowing state updates to be customized via setConfig, though not directly in ClientAgent.
CC_DEFAULT_STATE_GET: <T = any>(clientId: string, stateName: string, defaultState: T) => Promise<T>
Default function to get state values, used in IState.getState for state retrieval.
Returns defaultState by default, allowing state retrieval to be customized via setConfig, though not directly in ClientAgent.
CC_DEFAULT_POLICY_GET_BAN_CLIENTS: (policyName: string, swarmName: string) => string[] | Promise<string[]>
Default function to get banned clients for the policy
CC_DEFAULT_POLICY_GET: (policyName: string, swarmName: string) => string[] | Promise<string[]>
Retrieves the list of currently banned clients under this policy.
CC_DEFAULT_POLICY_SET: (clientIds: string[], policyName: string, swarmName: string) => void | Promise<void>
Optional function to set the list of banned clients. Overrides default ban list management if provided.
CC_DEFAULT_STORAGE_GET: <T extends IStorageData = IStorageData>(clientId: string, storageName: string, defaultValue: T[]) => Promise<T[]>
Default function to get storage data, used in IStorage.take for storage retrieval.
Returns defaultValue by default, allowing storage retrieval to be customized via setConfig, though not directly in ClientAgent.
CC_DEFAULT_STORAGE_SET: <T extends IStorageData = IStorageData>(data: T[], clientId: string, storageName: string) => Promise<void>
Default function to set storage data, used in IStorage.upsert for storage persistence.
No-op by default, allowing storage updates to be customized via setConfig, though not directly in ClientAgent.
CC_SKIP_POSIX_RENAME: boolean
Flag to skip POSIX-style renaming, potentially for file operations in persistence layers.
Disabled (false) by default, ensuring standard renaming unless overridden, though not directly in ClientAgent.
CC_PERSIST_MEMORY_STORAGE: boolean
Flag to enable persistent storage for Schema.readValue and Schema.writeValue, separate from general persistence.
Enabled (true) by default, ensuring memory storage persistence unless overridden.
CC_PERSIST_EMBEDDING_CACHE: boolean
Flag to enable persistent cache for embeddings. Will allow to reduce costs while using openai
Disabled (false) by default which faster for ollama local embeddings
CC_DEFAULT_READ_EMBEDDING_CACHE: (embeddingName: string, stringHash: string) => number[] | Promise<number[]>
Retrieves the embedding vector for a specific string hash, returning null if not found. Used to check if a precomputed embedding exists in the cache.
CC_DEFAULT_WRITE_EMBEDDING_CACHE: (embeddings: number[], embeddingName: string, stringHash: string) => void | Promise<void>
Stores an embedding vector for a specific string hash, persisting it for future retrieval. Used to cache computed embeddings to avoid redundant processing.
CC_DEFAULT_AGENT_TOOL_VALIDATE: (dto: { clientId: string; agentName: string; toolCalls: IToolCall[]; params: unknown; }) => boolean | Promise<boolean>
Validates the tool parameters before execution. Can return synchronously or asynchronously based on validation complexity.
CC_THROW_WHEN_NAVIGATION_RECURSION: boolean
Throw an error if agents being changed recursively