Reference

Getting Started

Core classes for building AI assistants

ChatBot: Main entry point for building and managing conversational AI chatbots with integrated
Conversation: Manages sequences of messages in conversational AI interactions, primarily returned by ChatBot methods.
Message: Represents a single message in a conversational AI interaction, discovered through Conversation objects.

ChatBot Methods

Configuration and interaction methods for the ChatBot class

ChatBot.model(): Configure the language model to use for generating responses.
ChatBot.provider_model(): Set provider and model using a single string (e.g., “openai:gpt-4o”).
ChatBot.preset(): Apply a pre-configured behavior template to instantly specialize the chatbot.
ChatBot.temperature(): Control the randomness and creativity level of chatbot responses.
ChatBot.max_tokens(): Set the maximum number of tokens for chatbot responses.
ChatBot.persona(): Set a persona that shapes the chatbot’s tone and behavior.
ChatBot.system_prompt(): Set a custom system prompt, overriding any preset system prompt.
ChatBot.get_system_prompt(): Get the final constructed system prompt that will be sent to the LLM.
ChatBot.avoid(): Configure topics or behaviors the chatbot must refuse to engage with.
ChatBot.get_avoid_topics(): Get the list of topics or behaviors this chatbot is configured to avoid.
ChatBot.tools(): Configure tools for this chatbot with unified API for custom and built-in tools.
ChatBot.add_tools(): Add specific built-in tools.
ChatBot.verbose(): Enable or disable verbose diagnostic output during chat interactions.
ChatBot.chat(): Send a message to the chatbot and get a response within a conversation context.
ChatBot.start_conversation(): Start a new conversation with this chatbot.
ChatBot.continue_conversation(): Continue an existing conversation with a new message.
ChatBot.show(): Display diagnostic information or launch chat interfaces.
ChatBot.get_config_summary(): Get a comprehensive summary of the current chatbot configuration.
ChatBot.get_config(): Return a copy of the full chatbot configuration dictionary.
ChatBot.prompt_builder(): Create an attention-optimized prompt builder for declarative prompt composition.
ChatBot.structured_prompt(): Configure the chatbot with a structured prompt built from keyword sections.
ChatBot.chain_prompts(): Chain multiple structured prompts in attention-optimized order.
ChatBot.check_llm_status(): Check the status of LLM integration and get setup help if needed.
ChatBot.quick_start(): Get a quick-start guide tailored to this chatbot’s current configuration.
ChatBot.create_chat_session(): Create a chat session backed by the configured LLM provider.

Prompt Engineering

Attention-optimized prompt construction

PromptBuilder: Builds structured prompts using attention mechanisms and cognitive principles.
PromptBuilder.persona(): Set a behavioral persona to anchor the model’s response style and establish expertise
PromptBuilder.task_context(): Define the primary task context that establishes what needs to be accomplished.
PromptBuilder.focus_on(): Set the primary focus that leverages both front-loading and recency bias for maximum
PromptBuilder.critical_constraint(): Add a critical constraint that will be front-loaded for maximum attention and impact.
PromptBuilder.constraint(): Add a standard constraint to the prompt that will appear in the additional constraints
PromptBuilder.core_analysis(): Define core analysis requirements as a high-priority, required structured section.
PromptBuilder.structured_section(): Add a structured section with clear hierarchical boundaries and visual organization.
PromptBuilder.avoid_topics(): Specify topics or behaviors to avoid through negative constraints that guide AI responses
PromptBuilder.output_format(): Specify output formatting requirements to prevent ambiguous responses and ensure structured
PromptBuilder.example(): Add an input/output example for few-shot learning and response format demonstration.
PromptBuilder.final_emphasis(): Set final emphasis that leverages recency bias to ensure critical instructions receive
PromptBuilder.vocabulary(): Add domain-specific vocabulary definitions to ensure consistent understanding of
PromptBuilder.pathways(): Add conversational pathway guidance to structure and guide conversation flow.
PromptBuilder.print(): Print the assembled prompt to stdout.
Priority: Control where a prompt section is placed relative to other sections.
PromptSection: Represents a structured section of an attention-optimized prompt with priority and ordering
VocabularyTerm: Define domain-specific terminology with multilingual support for consistent AI understanding.
architectural_analysis_prompt(): Create a pre-configured PromptBuilder for architectural analysis.
code_review_prompt(): Create a pre-configured PromptBuilder for code review.
debugging_prompt(): Create a pre-configured PromptBuilder for debugging.

Conversation Pathways

State-machine conversation flows with branching and fallbacks

Pathways: Chainable builder for defining structured conversational pathways.
Pathways.state(): Define a state with natural language description as the primary identifier.
Pathways.required(): Specify required information for the current state to be considered complete.
Pathways.optional(): Specify optional information that would be helpful but not required.
Pathways.tools(): Specify tools available for use in the current state.
Pathways.success_condition(): Define what indicates successful completion of the current state.
Pathways.next_state(): Define direct transition to the next state.
Pathways.branch_on(): Define conditional branch to another state based on specific conditions.
Pathways.fallback(): Define fallback transition when normal state progression fails.
Pathways.agent(): Assign an agent to the current state.
Pathways.register_agent(): Register an Agent instance for use in pathway states.
Pathways.get_agent(): Get the agent assigned to a specific state.
Pathways.visualize(): Create an HTML visualization of this pathway and save to file.

Conversation Methods

Methods for managing conversation history

Conversation.add_message(): Append a message with the given role to the conversation.
Conversation.add_user_message(): Add a user-role message to the conversation.
Conversation.add_assistant_message(): Add an assistant-role message to the conversation.
Conversation.add_system_message(): Add a system-role message to the conversation.
Conversation.get_messages(): Return messages, optionally filtered by role.
Conversation.get_last_message(): Return the most recent message, optionally filtered by role.
Conversation.clear_messages(): Remove all messages from the conversation.
Conversation.set_context_window(): Limit how many recent messages get_context_messages() returns.
Conversation.get_context_messages(): Return messages within the configured context window.
Conversation.get_message_count(): Return the total number of messages in the conversation.
Conversation.to_dict(): Serialize the conversation to a plain dictionary.
Conversation.from_dict(): Reconstruct a conversation from a dictionary produced by to_dict().

File Attachments

Multi-modal file handling for conversations

Attachments: File attachment handler for Talk Box conversations.
Attachments.with_prompt(): Add a text prompt to accompany the file attachments.
Attachments.to_chat_contents(): Convert attachments to chatlas-compatible content list.
Attachments.get_metadata(): Get metadata for all processed attachments.
Attachments.summary(): Get a human-readable summary of attached files.
AttachmentMetadata: Metadata for individual file attachments.

Tools

Tool registration, execution, and built-in utilities

tool(): Register a function as a Talk Box tool.
ToolCategory: Categorize tools by their primary function.
ToolContext: Provide runtime context to every tool execution.
ToolResult: Represent the outcome of a tool execution.
get_global_registry(): Return the process-wide ToolRegistry singleton.
load_tool_box(): Register all built-in Tool Box tools in the global registry.
get_builtin_tool(): Get a specific built-in tool by name.
load_selected_tools(): Register a subset of the built-in Tool Box tools.

Testing & Validation

Automated testing for chatbot quality assurance

autotest_avoid_topics(): Comprehensive avoid topics testing with automated violation detection.
autotest_pathways(): Automated testing for pathway adherence in chatbots.
TestResults: Enhanced test results container with rich reporting and analysis capabilities.
TestResults.to_dataframe(): Convert test results to a pandas DataFrame.
TestResults.to_great_table(): Create a styled Great Tables report from the test results.
PathwayTestResults: Aggregate and summarize the outcomes of a autotest_pathways() run.
ToolEnabledConversation: Extend Conversation with automatic tool invocation support.
create_tool_conversation(): Create a ToolEnabledConversation with default settings.

Guardrails

Composable input/output validation for chatbot messages

Guard: A configured guardrail ready to be attached to a ChatBot.
Guard.check(): Run the guard against the given text.
GuardAction: The outcome of a guard check.
GuardPhase: When the guard runs in the message lifecycle.
GuardResult: The result of evaluating a guardrail against a message.
GuardPipeline: Ordered pipeline of guards that processes messages sequentially.
GuardPipeline.add(): Add a guard to the end of the pipeline.
GuardPipeline.run(): Run all applicable guards against the text.
GuardPipelineResult: Result of running a message through the full guard pipeline.
guardrail(): Decorator that turns a function into a composable guardrail.
ChatBot.guardrail(): Add a guardrail to the chatbot’s validation pipeline.
ChatBot.guard_stats(): Get activation statistics for all attached guardrails.
no_pii(): Guard that detects and handles Personally Identifiable Information.
max_response_length(): Guard that enforces a maximum response length.
max_input_length(): Guard that rejects user messages exceeding a character limit.
tone_check(): Guard that flags responses not matching the expected tone.
disclaimer_required(): Guard that ensures a required disclaimer appears in the response.
must_cite_sources(): Guard that requires responses to contain source citations.
keyword_block(): Guard that blocks messages containing any of the specified keywords.

Evaluation

Prompt evaluation, scoring, and regression detection

eval.HAS_PANDAS
eval.HAS_GREAT_TABLES
eval.DEFAULT_DIMENSIONS
eval.ScorerFn
eval_regression(): Compare two bot versions and flag regressions.
eval_model_update(): Compare persona behavior across two model versions.
eval_suite(): Evaluate a persona across multiple models (model comparison matrix).
EvalCase: A single evaluation test case.
EvalDimension: Scoring dimensions for evaluation.
EvalResults: Collection of evaluation results with reporting capabilities.
EvalResults.summary(): Compute summary statistics for the eval run.
EvalResults.passed(): Check if all variants meet the minimum threshold.
EvalResults.regressions(): Detect regressions between variants.
EvalResults.scores_by_variant(): Aggregate mean scores per variant per dimension.
EvalResults.to_dataframe(): Export results to a pandas DataFrame.
EvalResults.to_great_table(): Create a Great Tables comparison report.
EvalResults.to_scorecard(): Export results as a scorecard dictionary (optionally written to JSON).
EvalResult: Results for a single query evaluated against a single bot variant.
EvalScore: A score on a single dimension for a single response.
scorecard_table(): Render a scorecard as a polished Great Table.
sweep_table(): Render an eval sweep summary as a polished Great Table.

Model Profiles

Model capability registry for routing, eval, and prompt optimization.

ModelProfile: Capability profile for a specific LLM model.
ModelProfile.key: Canonical provider:model key.
ModelProfile.name: Display name, falling back to model identifier.
ModelProfile.supports(): Check whether a capability is supported.
CostTier: Relative cost tier for a model.
get_model_profile(): Look up a model’s capability profile.
list_models(): List known model profiles, optionally filtered.
register_model(): Register a custom model profile.
model_profiles_table(): Render all known model profiles as a Great Table.

Ollama Detection

Auto-detect local Ollama instances and register available models.

OllamaStatus: Status of the local Ollama instance.
detect_ollama(): Detect whether a local Ollama instance is running.
list_ollama_models(): Query Ollama for available models and return their profiles.
sync_ollama_models(): Detect Ollama models and register them in the profile registry.

Context Window Management

Token-aware fitting of prompts and messages within model context limits.

ContextWindow: Manages fitting content into a model’s token budget.
ContextWindow.fit_messages(): Fit a list of conversation messages into the token budget.
ContextWindow.fit_prompt(): Fit a PromptBuilder’s output into the token budget.
ContextWindow.count_tokens(): Count tokens in text using the configured counter.
ContextWindow.fits(): Check whether text fits within the input budget.
ContextWindow.overflow(): Calculate how many tokens over budget the text is.
FitResult: Result of fitting content into a context window budget.
PromptFitResult: Result of fitting a PromptBuilder output into a token budget.
FitStrategy: Strategy for fitting content into a token budget.
estimate_tokens(): Estimate the token count for a string using a character-based heuristic.

Prompt Optimization

Reduce prompt token usage without losing semantic meaning.

optimize_prompt(): Optimize a prompt for reduced token usage.
OptimizationLevel: How aggressively to compress prompts.
OptimizeResult: Result of prompt optimization.

Hybrid Routing

Route tasks between local and cloud models based on complexity and cost.

route(): Route a task to the best available model.
classify_complexity(): Estimate the complexity of a task from its prompt text.
Router: Configurable model router for hybrid local/cloud routing.
Router.route(): Route a task to the best available model.
RoutingStrategy: Strategy that governs how the router selects models.
RoutingResult: Result of routing a task to a model.
TaskComplexity: Estimated complexity of a task or prompt.

Consensus Mode

Query multiple models, compare outputs, and resolve disagreements.

consensus: Consensus mode: query multiple models, compare outputs, flag disagreements.
find_disagreements(): Detect disagreements between model responses.
ConsensusStrategy: Strategy for resolving consensus across multiple model responses.
ConsensusResult: Result of running consensus across multiple model responses.
ModelResponse: A single model’s response to a prompt.
Disagreement: A detected disagreement between model responses.

Cascade Consensus

Start with one model, fan out to additional models when confidence is low.

cascade.Responder
estimate_confidence(): Estimate confidence in a model response based on textual signals.
CascadeResult: Result of a cascade consensus execution.
CascadeRound: Record of a single round in a cascade execution.

Memory Tiers

Working, short-term, and long-term memory for conversations.

MemoryStore: Unified memory interface across all three tiers.
MemoryStore.remember(): Store a memory in the specified tier.
MemoryStore.recall(): Retrieve a memory, searching tiers if no specific tier is given.
MemoryStore.forget(): Remove a memory from the specified tier, or from all tiers.
MemoryStore.search(): Search memories across tiers.
WorkingMemory: In-conversation key-value memory, lost when the conversation ends.
ShortTermMemory: Recent session memory with TTL and max-entry eviction.
LongTermMemory: Persistent memory backed by SQLite.
MemoryEntry: A single memory entry across any tier.
MemoryTier: Which memory tier an entry belongs to.

WorkingMemory Methods

Methods for the WorkingMemory class

WorkingMemory.set(): Store a value in working memory.
WorkingMemory.get(): Retrieve a value from working memory.
WorkingMemory.has(): Check if a key exists in working memory.
WorkingMemory.delete(): Remove a key from working memory.
WorkingMemory.keys(): List all keys in working memory.
WorkingMemory.entries(): List all entries in working memory.
WorkingMemory.search(): Search working memory by tags.
WorkingMemory.clear(): Remove all entries from working memory.
WorkingMemory.__len__()

ShortTermMemory Methods

Methods for the ShortTermMemory class

ShortTermMemory.set(): Store a value in short-term memory.
ShortTermMemory.get(): Retrieve a value from short-term memory.
ShortTermMemory.has(): Check if a key exists and is not expired.
ShortTermMemory.delete(): Remove a key from short-term memory.
ShortTermMemory.keys(): List all non-expired keys.
ShortTermMemory.entries(): List all non-expired entries.
ShortTermMemory.search(): Search short-term memory by tags.
ShortTermMemory.clear(): Remove all entries.
ShortTermMemory.__len__()

LongTermMemory Methods

Methods for the LongTermMemory class

LongTermMemory.set(): Store a value in long-term memory.
LongTermMemory.get(): Retrieve a value from long-term memory.
LongTermMemory.has(): Check if a key exists in long-term memory.
LongTermMemory.delete(): Remove a key from long-term memory.
LongTermMemory.keys(): List all keys in long-term memory.
LongTermMemory.entries(): List all entries in long-term memory.
LongTermMemory.search(): Search long-term memory by tags and/or key prefix.
LongTermMemory.clear(): Remove all entries from long-term memory.
LongTermMemory.close(): Close the database connection.
LongTermMemory.__len__()

Agent

Self-contained AI agent bundling persona, memory, tools, and conversation history.

Agent: A self-contained AI agent bundling persona, memory, and conversation history.
Agent.from_persona(): Create an agent from a registered persona name.
Agent.respond(): Send a message and get a response from this agent.
Agent.remember(): Store a value in the agent’s memory.
Agent.recall(): Retrieve a value from the agent’s memory.
Agent.retention: The agent’s retention policy (from its persona), if any.
Agent.conversation: The agent’s current conversation, if any.
Agent.reset_conversation(): Clear the agent’s conversation state and start fresh.

Agent Methods

Methods for the Agent class

Agent.__post_init__()
Agent.from_persona(): Create an agent from a registered persona name.
Agent.respond(): Send a message and get a response from this agent.
Agent.remember(): Store a value in the agent’s memory.
Agent.recall(): Retrieve a value from the agent’s memory.
Agent.reset_conversation(): Clear the agent’s conversation state and start fresh.

Human-in-the-Loop

Pause agent workflows for human review, approval, or revision.

human_review(): Create a human-in-the-loop review request.
approve(): Approve a pending review.
reject(): Reject a pending review.
revise(): Revise the content and approve the review.
HumanReview: A request for human review before proceeding.
HumanReview.status: Current status of this review.
HumanReview.decision: The human’s decision, or None if still pending.
HumanReview.is_pending: Whether the review is still awaiting a decision.
HumanReview.is_resolved: Whether the review has been decided (approved, rejected, or revised).
HumanReview.is_approved: Whether the review was approved (including revised).
HumanReview.final_content: The content to use going forward.
ReviewDecision: A human’s decision on a review request.
ReviewStatus: Status of a human-in-the-loop review.
ReviewQueue: Collects and tracks multiple review requests.

ReviewQueue Methods

Methods for the ReviewQueue class

ReviewQueue.add(): Add a review to the queue.
ReviewQueue.pending(): Return all pending reviews.
ReviewQueue.resolved(): Return all resolved reviews.
ReviewQueue.by_agent(): Return reviews from a specific agent.
ReviewQueue.by_status(): Return reviews matching a specific status.
ReviewQueue.__len__()

Skill System

Loadable capability packs for agents.

SkillDefinition: A loadable skill pack that adds capabilities to an agent.
create_skill(): Create a skill definition programmatically.
get_skill(): Get a skill definition by name.
list_skills(): List all available skill names.
register_skill(): Register a skill definition in the global registry.
load_skill(): Load a skill from an arbitrary YAML file.
skill_categories(): Get all skills grouped by category.

Structured Outputs

Extract typed data from LLM responses using Pydantic models.

ExtractResult: Result of extracting structured data from an LLM response.
extract(): Extract structured data from an LLM response.
schema_to_dict(): Convert a Pydantic model class to a JSON Schema dictionary.

Persona Traits

Composable modifiers that add constraints, expertise, and tools to any persona.

TraitDefinition: A composable modifier that can be applied to any persona.
apply_trait(): Apply a trait to a persona, returning a new modified persona.
create_trait(): Create a trait definition programmatically.
get_trait(): Get a registered trait by name.
list_traits(): List all available trait names.
register_trait(): Register a trait so it can be retrieved with get_trait().
load_trait(): Load a trait from an arbitrary YAML file path.
trait_categories(): Get all traits grouped by category.

MCP Bridge

Bidirectional bridge between Talk Box tools and the Model Context Protocol.

MCPToolInfo: Description of an MCP tool discovered from a remote server.
MCPBridgeServer: An MCP server that exposes Talk Box tools over the MCP protocol.
MCPBridgeServer.tool_names(): Return the names of all tools registered with this server.
MCPBridgeServer.build(): Build and return a FastMCP server instance.
MCPBridgeServer.run(): Build the server (if needed) and run it.
tools_to_mcp_server(): Create an MCPBridgeServer from a list of Talk Box tools.
mcp_tool_to_talk_box(): Convert an MCPToolInfo into a TalkBoxTool.
list_mcp_tools(): Convert raw MCP tool dicts into MCPToolInfo objects.
discover_mcp_tools(): Discover tools from a live MCP client session.

Knowledge Graph

SQLite-backed knowledge graph with nodes, edges, traversal, search, and embedding support.

KnowledgeGraph: SQLite-backed knowledge graph with nodes, edges, and optional embeddings.
KnowledgeGraph.add_node(): Add a node to the graph, or update it if the ID already exists.
KnowledgeGraph.get_node(): Get a node by ID.
KnowledgeGraph.delete_node(): Delete a node and all its edges.
KnowledgeGraph.list_nodes(): List nodes, optionally filtered by type.
KnowledgeGraph.node_count(): Count nodes in the graph.
KnowledgeGraph.add_edge(): Add an edge to the graph, or update it if the key already exists.
KnowledgeGraph.get_edges(): Get edges connected to a node.
KnowledgeGraph.delete_edge(): Delete a specific edge.
KnowledgeGraph.edge_count(): Count edges in the graph.
KnowledgeGraph.neighbors(): Get neighboring nodes connected by edges.
KnowledgeGraph.search(): Search nodes by name or content (case-insensitive substring match).
KnowledgeGraph.stats(): Get summary statistics about the graph.
KnowledgeGraph.health(): Get health metrics for the graph.
KnowledgeGraph.pending_questions(): Get pending enrichment questions for this knowledge graph.
KnowledgeGraph.answer_question(): Answer a pending enrichment question.
KnowledgeGraph.dismiss_question(): Dismiss a pending enrichment question without answering.
KnowledgeGraph.question_stats(): Get summary statistics for the enrichment question queue.
KnowledgeGraph.clear(): Delete all nodes and edges from the graph.
KnowledgeGraph.close(): Close the database connection.
Node: A node in the knowledge graph.
Edge: A directed relationship between two nodes.
NodeType: Type of a knowledge graph node.
cosine_similarity(): Compute cosine similarity between two vectors.

KnowledgeGraph Methods

Methods for the KnowledgeGraph class

KnowledgeGraph.add_node(): Add a node to the graph, or update it if the ID already exists.
KnowledgeGraph.get_node(): Get a node by ID.
KnowledgeGraph.delete_node(): Delete a node and all its edges.
KnowledgeGraph.list_nodes(): List nodes, optionally filtered by type.
KnowledgeGraph.node_count(): Count nodes in the graph.
KnowledgeGraph.add_edge(): Add an edge to the graph, or update it if the key already exists.
KnowledgeGraph.get_edges(): Get edges connected to a node.
KnowledgeGraph.delete_edge(): Delete a specific edge.
KnowledgeGraph.edge_count(): Count edges in the graph.
KnowledgeGraph.neighbors(): Get neighboring nodes connected by edges.
KnowledgeGraph.search(): Search nodes by name or content (case-insensitive substring match).
KnowledgeGraph.stats(): Get summary statistics about the graph.
KnowledgeGraph.health(): Get health metrics for the graph.
KnowledgeGraph.pending_questions(): Get pending enrichment questions for this knowledge graph.
KnowledgeGraph.answer_question(): Answer a pending enrichment question.
KnowledgeGraph.dismiss_question(): Dismiss a pending enrichment question without answering.
KnowledgeGraph.question_stats(): Get summary statistics for the enrichment question queue.
KnowledgeGraph.clear(): Delete all nodes and edges from the graph.
KnowledgeGraph.close(): Close the database connection.
KnowledgeGraph.__repr__()

KG Visualization

Render knowledge graphs as Mermaid diagrams or interactive HTML pages.

VisNode: A node prepared for visualization.
VisEdge: An edge prepared for visualization.
VisGraph: A complete graph prepared for visualization.
VisGraph.node_count: Number of nodes in the visualization.
VisGraph.edge_count: Number of edges in the visualization.
VisGraph.to_dict(): Serialize to a dictionary.
prepare_vis_graph(): Prepare a knowledge graph for visualization.
to_mermaid(): Render a knowledge graph as a Mermaid diagram string.
to_html(): Render a knowledge graph as an interactive HTML page.
visualize(): Render a knowledge graph and open it in the browser.

Connectors

Ingest documents from files, directories, and Apple Notes into the knowledge graph.

Document: A document to ingest into the knowledge graph.
Connector: Base class for knowledge graph connectors.
Connector.scan(): Yield documents from this source.
MarkdownDir: Ingest Markdown files from a directory.
MarkdownDir.scan(): Yield Markdown documents from the directory tree.
DirectoryConnector: Ingest text-based files from a directory by extension.
DirectoryConnector.scan(): Yield documents from the directory tree.
AppleNotes: Ingest notes from macOS Apple Notes (read-only).
AppleNotes.scan(): Yield documents from Apple Notes matching the folder/tag filter.
SyncResult: Result of a connector sync operation.
SyncResult.total: Total documents processed.
connector(): Create a connector from a generator function.
sync(): Sync documents from connectors into a knowledge graph.

Enrichment

Extract entities, topics, relationships, and summaries from documents in the knowledge graph.

ExtractedEntity: An entity extracted from a document.
ExtractedRelationship: A relationship extracted between two entities.
EnrichmentResult: Complete enrichment output for a single document.
EnrichmentResult.entity_names: Names of all extracted entities.
EnrichmentResult.entity_count: Number of extracted entities.
EnrichmentResult.topic_count: Number of assigned topics.
EnrichmentResult.relationship_count: Number of extracted relationships.
EnrichmentPipeline: Configurable pipeline for enriching documents in a knowledge graph.
EnrichmentPipeline.run(): Run enrichment on document nodes in the knowledge graph.
PipelineResult: Result of running the enrichment pipeline.
PipelineResult.total: Total documents processed (enriched + skipped).
EnrichmentConfig: Configuration for how enrichment results are applied.
regex_enricher(): Simple regex-based enrichment (no LLM required).

Compliance Export

Export conversation transcripts as auditable JSON or HTML files.

export_json(): Export a conversation capture as an auditable JSON file.
export_html(): Export a conversation capture as a human-readable HTML transcript.

CLI

Command-line interface for managing and testing AI assistants.

cli.main(): Talk Box: build production AI assistants.
cli.info(): Show Talk Box configuration, personas, and models.
cli.personas(): List personas or show details for a specific persona.
cli.models(): List registered model profiles.
cli.test_cmd(): Run eval suite for a persona across one or more models.

Configuration

Layered configuration system with profiles, environment overrides, and CLI flags.

TalkBoxConfig: Complete Talk Box configuration.
TalkBoxConfig.resolve(): Resolve effective settings from config + profile + overrides.
TalkBoxConfig.to_dict()
TalkBoxConfig.validate_model(): Validate that a model string is allowed under the current config.
TalkBoxConfig.is_cloud_model(): Check whether a model string refers to a cloud provider.
ProfileConfig: A named profile bundling model + persona + settings.
ProfileConfig.to_dict()
ResolvedConfig: The final resolved configuration after all layers are merged.
TUIMode: TUI display mode.
load_config(): Load and merge the full config stack.
save_config(): Write a config to a YAML file.
load_profile(): Load a named profile from the profiles directory.
save_profile(): Save a named profile to the profiles directory.
list_profiles(): List available profile names from the profiles directory.