Glossary

---

title: "Glossary" tier: "reference" status: "verified" version: "1.2.0" last-verified: "2026-06-17" commit: "b1c347b1" network-badge: "none" risk-tags: []

Important

AI Assist Note (Knowledge Heritage): This document is part of the "Sovereign Reality" documentation.

• @docs ARCHITECTURE:Documentation
• Failure Path: Information drift, legacy terminology, or documentation mismatch.
• Telemetry Link: Search [GLOSSARY] in audit logs.

AI Assist Note

📖 Tadpole OS: Technical Specifications & Glossary

🔍 Debugging & Observability

Traceability via execution/parity_guard.py.

📖 Tadpole OS: Technical Specifications & Glossary

Intelligence Level: High (ECC Optimized) Source of Truth: Rust Source Code (server-rs/src/) Standard Compliance: ECC-SPEC-01 (Strict Reference Protocol)

---

🏗️ Constants & Statics

TELEMETRY_TX

[Tier 3: Security] Type: static Purpose: Global broadcast channel for high-throughput JSON telemetry emissions from system spans and agent lifecycle events. Parameters: N/A Return Value: Lazy<broadcast::Sender<serde_json::Value>>. The sender side of a tokio broadcast channel with a capacity of 2000 events. Side Effects: N/A Failure Conditions: N/A

---

🏗️ Structs

AppState

[Tier 2: Admin] Type: struct Purpose: Primary global application state container and coordination hub for all sovereign subsystems. Parameters:

• comms (Arc<CommunicationHub>): Definition: Manages real-time communication channels and telemetry.
• governance (Arc<GovernanceHub>): Definition: Manages operational limits and global policy settings.
• registry (Arc<RegistryHub>): Definition: Maintains thread-safe registries for agents, providers, and skills.
• security (Arc<SecurityHub>): Definition: Handles audit trails, budget enforcement, and shell safety scanning.
• resources (Arc<ResourceHub>): Definition: Manages shared system resources including database pools and HTTP clients.
• base_dir (PathBuf): Definition: The root directory for persistent data and workspace sandboxes. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

CommunicationHub

[Tier 2: Admin] Type: struct Purpose: Orchestrates real-time event broadcasting and human-in-the-loop oversight resolution. Parameters:

• tx (broadcast::Sender<LogEntry>): Definition: Broadcast system logs to all connected UI WebSockets.
• event_tx (broadcast::Sender<serde_json::Value>): Definition: Dedicated broadcast for Engine events (decisions, lifecycle changes).
• telemetry_tx (broadcast::Sender<serde_json::Value>): Definition: Dedicated high-speed broadcast for agent telemetry (thinking, status).
• audio_stream_tx (broadcast::Sender<Vec<u8>>): Definition: Dedicated high-speed broadcast for neural audio streams (PCM chunks).
• pulse_tx (broadcast::Sender<Arc<crate::telemetry::pulse_types::SwarmPulse>>): Definition: High-speed binary pulse broadcasting for swarm visualization.
• oversight_queue (DashMap<String, OversightEntry>): Definition: Pending Oversight entries awaiting human decision.
• oversight_resolvers (DashMap<String, oneshot::Sender<crate::agent::types::OversightResolution>>): Definition: Resolvers for pending oversight promises.
• active_runners (DashMap<String, tokio::task::AbortHandle>): Definition: Active AbortHandles for running agents, allowing for definitive task cancellation.
• runner_semaphore (tokio::sync::Semaphore): Definition: Semaphore to limit concurrent executing agents (runner pool throttle).
• event_sequence (std::sync::atomic::AtomicU64): Definition: Monotonic sequence counter for outbound engine events. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

GovernanceHub

[Tier 2: Admin] Type: struct Purpose: Centralizes system limits and automated policy enforcement for the agent swarm. Parameters:

• auto_approve_safe_skills (AtomicBool): Definition: Global setting: whether to auto-approve low-risk skills.
• max_agents (AtomicU32): Definition: Maximum allowed agents in the swarm (Default: 50).
• max_clusters (AtomicU32): Definition: Maximum allowed concurrent mission clusters permitted (Default: 10).
• max_swarm_depth (AtomicU32): Definition: Maximum depth for agent recursion/spawning (Default: 5).
• max_task_length (AtomicUsize): Definition: Maximum token length for a single task (Default: 32768).
• default_budget_usd (RwLock<f64>): Definition: Default budget allocated to new agents in USD (Default: 1.0 USD).
• active_agents (AtomicU32): Definition: Number of agents currently executing tasks.
• recruit_count (AtomicU32): Definition: Total number of recruitment operations performed.
• tpm_accumulator (AtomicUsize): Definition: Global TPM (Tokens Per Minute) accumulator for telemetry.
• privacy_mode (AtomicBool): Definition: Privacy Shield: When true, all external cloud provider traffic is blocked.
• failover_amber_threshold (AtomicU32): Definition: Failover Amber threshold (failures before status becomes Amber, Default: 3).
• failover_red_threshold (AtomicU32): Definition: Failover Red threshold (failures before status becomes Red, Default: 5).
• failover_max_attempts (AtomicU32): Definition: Failover Max attempts (max retries to alternate models, Default: 3).
• provider_timeout_secs (AtomicU32): Definition: Default timeout for LLM provider generation calls in seconds (Default: 60).
• null_providers_test_mode (AtomicBool): Definition: Test mode flag to route all LLMs through NullProvider (TADPOLE_NULL_PROVIDERS=true).
• deprecated_routes (RwLock<std::collections::HashMap<String, (String, String)>>): Definition: Deprecated endpoints mapped to (Sunset Date, Alternate Link). Return Value: N/A Side Effects: N/A Failure Conditions: N/A

RegistryHub

[Tier 2: Admin] Type: struct Purpose: Centralized directory for swarm identities, model providers, and skill discovery. Parameters:

• agents (DashMap<String, EngineAgent>): Definition: The live agent registry, synced with persistence.
• providers (DashMap<String, ProviderConfig>): Definition: Configured LLM providers (e.g., OpenAI, Ollama).
• provider_health (DashMap<String, ProviderStatus>): Definition: Real-time health status of providers (Amber/Red state machine).
• provider_failures (DashMap<String, std::sync::atomic::AtomicU32>): Definition: Recent failure counts for providers to trigger health transitions.
• models (DashMap<String, ModelEntry>): Definition: Available LLM models catalog.
• nodes (DashMap<String, SwarmNode>): Definition: Discovery registry for infrastructure nodes in the swarm.
• skills (Arc<ScriptSkillsRegistry>): Definition: Registry for dynamic file-based Skills and Workflows.
• skill_registry (Arc<SkillRegistry>): Definition: Manager for dynamic Skill Manifests (skill.json).
• mcp_host (Arc<McpHost>): Definition: Host for Model Context Protocol (MCP) tool aggregation.
• hooks (Arc<HooksManager>): Definition: Manager for Lifecycle Hooks (Pre/Post tool execution).
• tool_registry (Arc<crate::agent::runner::tools::registry::ToolRegistry>): Definition: Unified registry for all builtin and categorical tools. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

ResourceHub

[Tier 2: Admin] Type: struct Purpose: Manages thread-safe access to heavy system resources and shared mission context. Parameters:

• pool (SqlitePool): Definition: SQLite connection pool for persistent storage.
• http_client (Arc<Client>): Definition: Shared HTTP client with optimized connection pooling.
• audio_engine (OnceCell<Arc<NeuralAudioEngine>>): Definition: Native engine for local audio synthesis (PCM) and transcription (loaded lazily).
• audio_cache (Arc<BunkerCache>): Definition: Zero-latency semantic audio replicate cache for frequent phrases.
• code_graph (OnceCell<Arc<RwLock<CodeGraph>>>): Definition: Graph of code relationships for RAG-enhanced tool search.
• symbol_graph (OnceCell<Arc<RwLock<CodeSymbolGraph>>>): Definition: Symbol-level Knowledge Graph.
• obfuscation_salt (String): Definition: Dynamic boot-time cryptographically secure salt. Used for path obfuscation.
• identity_context (OnceCell<String>): Definition: Global system identity context loaded from directives/IDENTITY.md.
• memory_context (OnceCell<String>): Definition: Global long-term memory context loaded from directives/LONG_TERM_MEMORY.md.
• swarm_vault (OnceCell<Arc<VectorMemory>>): Definition: Global swarm-wide knowledge vault for cross-mission intelligence (compiled when vector-memory feature is active).
• knowledge_store (OnceCell<Arc<crate::agent::knowledge_store::KnowledgeStore>>): Definition: Persistent cross-cluster Institutional Knowledge Store (compiled when vector-memory feature is active).
• rate_limiters (DashMap<String, Arc<RateLimiter>>): Definition: Cached rate limiters partitioned by model and provider.
• initialization_registry (DashMap<String, SubsystemStatus>): Definition: Tracks the initialization status of all subsystems.
• hardware_profiler (Arc<crate::system::profiler::HardwareProfiler>): Definition: System hardware profiler for sovereign compute telemetry.
• acl (Arc<dyn crate::agent::runner::service_traits::AclServiceTrait>): Definition: Global Access Control List service for tool governance.
• renderer (Arc<dyn crate::agent::runner::service_traits::PromptRendererTrait>): Definition: System prompt template renderer.
• base_dir (std::path::PathBuf): Definition: Base directory for relative path resolution.
• tool_cache (Arc<parking_lot::Mutex<crate::agent::runner::tools::cache::SharedToolCache>>): Definition: Shared cache for read-only tool results.
• conflict_manager (Arc<crate::security::conflict::ConflictManager>): Definition: Registry for conflict locks during concurrent file operations.
• blueprint_cache (tokio::sync::OnceCell<Arc<crate::services::blueprint_service::Blueprint>>): Definition: Cached codebase blueprint for System 2 indexing. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

SecurityHub

[Tier 3: Security] Type: struct Purpose: Manages tamper-evident auditing and preventative security protocols. Parameters:

• audit_trail (Arc<MerkleAuditTrail>): Definition: Tamper-evident audit trail engine (Merkle Hash Chain).
• budget_guard (Arc<BudgetGuard>): Definition: Persistent budget governance and metering engine.
• shell_scanner (Arc<ShellScanner>): Definition: Proactive shell safety scanner (API key leak protection).
• secret_redactor (Arc<SecretRedactor>): Definition: Runtime secret redactor for logs and telemetry.
• system_monitor (Arc<SecurityMonitor>): Definition: System resource and environment monitor.
• permission_policy (Arc<PermissionPolicy>): Definition: Dynamic tool permission and governance policy engine.
• deploy_token (String): Definition: Authentication token for administrative/deploy requests. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

LogEntry

[Tier 3: Security] Type: struct Purpose: Atomic unit of telemetry mirroring the frontend event interface. Parameters:

• event_type (String): Definition: Category of event (renamed to "type" in JSON, e.g. "log").
• id (String): Definition: Unique UUID v4 identifier.
• timestamp (DateTime<Utc>): Definition: Event creation time.
• source (String): Definition: Originating subsystem (e.g., "System", "Agent").
• severity (String): Definition: Log level (e.g., "INFO", "CRITICAL").
• agent_id (Option<String>): Definition: Optional ID of the agent associated with this log.
• agent_name (Option<String>): Definition: Optional name of the agent associated with this log.
• mission_id (Option<String>): Definition: Optional correlation ID for the swarm mission.
• text (String): Definition: The redacted event message. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

TelemetryLayer

[Tier 3: Security] Type: struct Purpose: Tracing subscriber layer bridging internal spans to OTel-compatible JSON events. Parameters:

• redactor (SecretRedactor): Definition: Redacts sensitive patterns within log messages. Return Value: N/A Side Effects: Broadcasts JSON events to TELEMETRY_TX on span creation and closure. Failure Conditions: N/A

AgentRunner

[Tier 2: Admin] Type: struct Purpose: Execution engine orchestrating the autonomous mission lifecycle. Parameters:

• state (Arc<AppState>): Definition: Reference to global application state.
• model_router (Arc<dyn ModelRouter>): Definition: Interface for routing LLM inference calls to appropriate providers.
• prompt_service (Arc<dyn PromptService>): Definition: Interface for loading and building prompt templates.
• tool_orchestrator (Arc<dyn ToolOrchestrator>): Definition: Interface for executing external tools with budget checks.
• mission_state_manager (Arc<dyn MissionStateManager>): Definition: Interface for saving/loading mission context.
• workflow_coordinator (Arc<dyn WorkflowCoordinator>): Definition: Coordinator for executing deterministic SOP sequences. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

RunContext

[Tier 2: Admin] Type: struct Purpose: Data container carrying mission state and environment throughout an execution loop. Parameters:

• agent_id (String): Definition: Unique identifier for the agent.
• name (String): Definition: Human-readable name of the agent.
• role (String): Definition: Dynamic role assigned to the agent.
• department (String): Definition: Functional department of the agent.
• description (String): Definition: Summary of the agent's responsibilities.
• model_config (ModelConfig): Definition: Primary LLM configuration.
• skills (Vec<String>): Definition: List of enabled custom files/skills.
• workflows (Vec<String>): Definition: List of SOP workflows registered to the agent.
• agent_models (AgentModels): Definition: Alternate slots for planning and execution LLMs.
• mcp_tools (Vec<String>): Definition: List of tool names exposed by active MCP servers.
• mission_id (String): Definition: Unique identifier for the current cluster mission.
• user_id (Option<String>): Definition: Optional identifier of the initiating user.
• depth (u32): Definition: Swarm recursion depth level (limit-controlled).
• lineage (Vec<String>): Definition: Parent agent IDs leading to this instance.
• provider_name (String): Definition: Name of the active LLM provider.
• workspace_root (PathBuf): Definition: Sandbox directory for file operations.
• fs_adapter (FilesystemAdapter): Definition: Zero-trust adapter restricting path traversals.
• safe_mode (bool): Definition: When enabled, hazardous commands/actions require user confirmation.
• analysis (bool): Definition: Toggle flag for deep tracing metrics.
• traceparent (Option<String>): Definition: OTel trace context propagation header.
• last_accessed_files (Arc<Mutex<Vec<String>>>): Definition: Dynamic tracking of files read during execution.
• modified_files (Arc<Mutex<Vec<String>>>): Definition: Dynamic tracking of files modified during execution.
• commands_run (Arc<Mutex<HashSet<String>>>): Definition: Set of external shell commands run by the agent.
• allowed_files (Option<Vec<String>>): Definition: Specific file whitelist constraints (if sandbox-configured).
• recent_findings (Option<String>): Definition: Cached discoveries/findings for cross-turn retention.
• working_memory (Value): Definition: Ephemeral JSON scratchpad for the agent's internal state.
• base_dir (PathBuf): Definition: Root persistence directory of the engine server.
• summarized_history (Option<String>): Definition: Compacted representation of past turns to reduce token window pressure.
• structured_output (bool): Definition: Toggle flag requesting structured JSON format.
• backlog (Option<Arc<Mutex<MissionBacklog>>>): Definition: Queue of pending subtasks generated during execution.
• primary_goal (Option<String>): Definition: The top-level mission goal string.
• budget_usd (f64): Definition: Monetary cap allocated for the run.
• current_cost_usd (f64): Definition: Total cost incurred during the active run (cumulative).
• reasoning_depth (u32): Definition: Number of steps used for reasoning/CoT.
• act_threshold (f32): Definition: Score threshold for taking actions.
• max_turns (u32): Definition: Maximum iterations allowed for the run loop.
• authority_level (RoleAuthorityLevel): Definition: Dynamic permission tier mapped to the agent's role.
• resource_weights (HashMap<String, f32>): Definition: Resource priority distribution parameters.
• graph_context (Option<String>): Definition: Codebase symbol relationships injected for target files.
• verification_passed (bool): Definition: Flag tracking testing verification results. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

ModelConfig

[Tier 2: Admin] Type: struct Purpose: Configuration for an LLM instance including system prompts and performance thresholds. Parameters:

• provider (ModelProvider): Definition: Target backend provider protocol.
• model_id (String): Definition: Specific model identifier.
• api_key (Option<String>): Definition: Optional access key override.
• base_url (Option<String>): Definition: Optional backend endpoint URL override.
• system_prompt (Option<String>): Definition: Custom system prompt instruction override.
• temperature (Option<f32>): Definition: LLM sampling temperature.
• max_tokens (Option<u32>): Definition: Maximum response tokens.
• external_id (Option<String>): Definition: Telemetry / tracking identifier.
• rpm (Option<u32>): Definition: Rate limit threshold (Requests Per Minute).
• rpd (Option<u32>): Definition: Rate limit threshold (Requests Per Day).
• tpm (Option<u32>): Definition: Rate limit threshold (Tokens Per Minute).
• tpd (Option<u32>): Definition: Rate limit threshold (Tokens Per Day).
• skills (Option<Vec<String>>): Definition: Enabled skill whitelists.
• workflows (Option<Vec<String>>): Definition: Enabled SOP workflow mappings.
• mcp_tools (Option<Vec<String>>): Definition: Enabled MCP tool selections.
• steering_vectors (Option<Vec<String>>): Definition: Model alignment steering configurations.
• reasoning_depth (Option<u32>): Definition: Maximum internal reasoning turns (e.g. for o1/r1 models).
• act_threshold (Option<f32>): Definition: Confidence threshold for autonomous actions.
• max_turns (Option<u32>): Definition: Maximum turn limit override.
• connector_configs (Option<Vec<ConnectorConfig>>): Definition: Database or knowledge connector details.
• extra_parameters (Option<HashMap<String, Value>>): Definition: Additional provider-specific parameters. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

ModelCapabilities

[Tier 2: Admin] Type: struct Purpose: Feature matrix defining the functional limits and supported modalities of an LLM. Parameters:

• supports_tools (bool): Definition: Support for external function calling.
• supports_vision (bool): Definition: Support for image processing.
• supports_structured_output (bool): Definition: Support for structured JSON schemas in responses.
• supports_reasoning (bool): Definition: Support for internal reasoning loops (CoT).
• supports_halting_tool (bool): Definition: Support for explicitly yielding/stopping execution turns.
• context_window (u32): Definition: Total token capacity.
• max_output_tokens (u32): Definition: Maximum generation limit per API call. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

ProblemDetails

[Tier 3: Security] Type: struct Purpose: RFC 9457 compliant structure for machine-readable API error responses. Parameters:

• type_uri (String): Definition: URI identifying the specific error type (formatted as https://tadpole.os/errors/{slug}).
• title (String): Definition: Human-readable summary of the error.
• status (u16): Definition: HTTP status code.
• detail (String): Definition: Sanitized and redacted explanation of the failure.
• instance (Option<String>): Definition: Unique request/transaction URI path context.
• error_code (Option<String>): Definition: Uppercase alphanumeric error code representation.
• help_link (Option<String>): Definition: Publicly accessible documentation link for troubleshooting.
• severity (String): Definition: Logging severity tier (e.g., "CRITICAL", "ERROR", "WARNING"). Return Value: N/A Side Effects: N/A Failure Conditions: N/A

InstallTemplateRequest

[Tier 2: Admin] Type: struct Purpose: Request payload for installing a swarm template from a remote repository. Parameters:

• repository_url (String): Definition: Git URL of the template source.
• path (String): Definition: Path within the repository for the template configuration. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

KnowledgeEntry

[Tier 1: Operator] Type: struct Purpose: Represents a single record of institutional memory stored in the Institutional Knowledge Store. Parameters:

• id (String): Definition: Unique UUID v4 identifying the entry.
• text (String): Definition: The raw content snippet.
• topic (String): Definition: Categorical classification or tag.
• cluster_id (Option): Definition: Optional cluster/project workspace context.
• source_node_id (Option): Definition: The Bunker node that authored this entry.
• source_agent_id (Option): Definition: The agent that authored this entry.
• content_hash (String): Definition: SHA-256 fingerprint for deduplication.
• confidence (f32): Definition: Accuracy score (0.0 to 1.0) decays over time unless human-confirmed.
• human_confirmed (bool): Definition: Flag indicating manual validation by an operator.
• ttl (Option): Definition: Expiration timestamp.
• created_at (i64): Definition: Creation timestamp.
• access_count (i64): Definition: Number of times this node has been read.
• concept_type (String): Definition: OKF concept classification.
• title (Option): Definition: Descriptive header.
• description (Option): Definition: Extended documentation.
• resource_uri (Option): Definition: Canonical target link.
• tags (Option): Definition: Comma-separated descriptors. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

AddKnowledgeRequest

[Tier 1: Operator] Type: struct Purpose: Payload parameters for submitting a new knowledge asset to the IKS. Parameters:

• text (String): Definition: Raw textual content of the knowledge node.
• topic (String): Definition: General category.
• cluster_id (Option): Definition: Optional cluster scope.
• source_node_id (Option): Definition: Local/remote node ID.
• source_agent_id (Option): Definition: Authoring agent ID.
• confidence (Option): Definition: Initial confidence score.
• ttl_days (Option): Definition: Number of days until expiry.
• human_confirmed (Option): Definition: Pre-confirm flag.
• concept_type (Option): Definition: OKF concept type.
• title (Option): Definition: Descriptive title.
• description (Option): Definition: Extended description.
• resource_uri (Option): Definition: Canonical reference link.
• tags (Option): Definition: Tags. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

---

🏗️ Enums

AppError

[Tier 3: Security] Type: enum Purpose: Unified application error enumeration for the engine, mapping failures to RFC 9457 / HTTP protocols. Variants:

• Runner(RunnerError): Wrapper for execution errors occurring in the agent runner.
• Skill(SkillError): Wrapper for validation, safety, or recruitment errors inside custom script skills.
• BadRequest(String): Malformed HTTP request or validation failure.
• Unauthorized(String): Missing or invalid authentication token.
• Forbidden(String): Action blocked by access control policies.
• NotFound(String): Requested resource or entity does not exist.
• DomainError { code, detail, help_link }: Business logic failure, utilizing DomainCode.
• InfrastructureError { provider_id, kind, detail, help_link }: Downstream provider service failure, utilizing ProviderId and InfrastructureErrorKind.
• QuantizationFallback { model_id, suggested_quant, detail }: Mismatch in local model resource allocation.
• NotImplemented(String): Action or protocol is not implemented.
• RateLimit(String): API or provider rate limits exceeded.
• InternalServerError(String): Unhandled internal system error.
• Conflict(String): Conflicting concurrent operations (e.g. file lock).
• DegradedState(String): Subsystem is in a degraded or unavailable state.
• MultiError(Vec<AppError>): Aggregated collection of multiple concurrent errors.
• Anyhow(anyhow::Error): Wildcard wrapper for custom contextual errors.
• Sqlx(sqlx::Error): Database persistence failure.
• Io(std::io::Error): Standard system input/output failure.
• Reqwest(reqwest::Error): Downstream HTTP client failure.
• Serde(serde_json::Error): Serialization/deserialization failure.
• WalkDir(walkdir::Error): Filesystem directory traversal failure.
• Graph(GraphError): Error occurring within the code structure or symbol knowledge graphs. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

RunnerError

[Tier 3: Security] Type: enum Purpose: Specific execution and reasoning loop errors wrapped by AppError. Variants:

• BudgetExhausted(String): Mission cost exceeds the allocated budget limit.
• RecursionBlocked(String): Swarm recursion depth limit has been reached.
• SentinelGate(String): Guardrails or sentinel checks blocked execution.
• Compression(String): Failure during agent turn history compression. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

SkillError

[Tier 3: Security] Type: enum Purpose: Specific custom script skill and recruitment errors wrapped by AppError. Variants:

• ValidationError(String): Input or configuration data failed validation checks.
• RecruitmentFailure { recipe_id, role, detail }: Spawning of a swarm sub-agent failed.
• SanitizationViolation(String): Whitelist security scanner blocked unsafe content or command execution. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

RoleAuthorityLevel

[Tier 2: Admin] Type: enum Purpose: Hierarchical authority of an agent within the swarm. Variants:

• Executive: Strategic oversight and delegation (CEO, Overlord).
• Management: Tactical coordination (Alpha Node).
• Specialist: Standard task execution.
• Observer: Read-only auditing. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

ModelProvider

[Tier 2: Admin] Type: enum Purpose: Supported backend protocols and hosting platforms for LLM interaction. Variants:

• Openai, Anthropic, Google, Gemini, Ollama, Groq, Mistral, Perplexity, Fireworks, Together, Deepseek, Xai, Inception, Openrouter, Cerebras, Sambanova, OllamaCloud. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

SubsystemStatus

[Tier 3: Security] Type: enum Purpose: Lifecycle state machine for core engine components. Variants:

• NotStarted: Component is uninitialized.
• Warming(f32): Component is warming up, carrying the progress float (0.0 to 1.0).
• Ready: Component is fully initialized and operational.
• Failed(String): Terminal initialization failure state. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

SystemHealthState

[Tier 3: Security] Type: enum Purpose: Overall status reporting for the swarm engine. Variants:

• Warming: Critical subsystems are starting up.
• Ready: All critical subsystems operational.
• Degraded: Subsystems failed or are in a degraded state. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

Modality

[Tier 2: Admin] Type: enum Purpose: Feature modality categories. Variants:

• Llm, Vision, Voice, Audio, Reasoning. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

MissionStatus

[Tier 1: Operator] Type: enum Purpose: Swarm mission lifecycle status. Variants:

• Pending, SpecReview, Active, Completed, Failed, Paused. Return Value: N/A Side Effects: N/A Failure Conditions: N/A

---

📖 Terminology & Concepts

Agent Scope

[Tier 2: Admin]

Note

Redirect: See RAG Scope for the canonical definition of Agent Scope / RAG Scope.

Alpha Node

[Tier 1: Operator]

Note

Renamed Term: This term has been renamed to Department Lead. See Department Lead for the canonical definition.

AXUM

[Tier 2: Admin] Type: Technical Framework Purpose: The high-performance, asynchronous Rust web framework used to run the Tadpole OS backend engine and expose HTTP/WebSocket service APIs. Implementation Hook: server-rs/src/router.rs

Circuit Breaker

[Tier 3: Security] Type: Terminology Purpose: Client-side resilience mechanism that monitors communication failures inside service namespaces (e.g. infra, engine) and short-circuits further requests to prevent dashboard locks. Implementation Hook: VITE_SYSTEM_BREAKER_FAILURE_THRESHOLD / VITE_SYSTEM_BREAKER_COOLDOWN_MS

Context Compaction

[Tier 2: Admin] Type: Terminology Purpose: The automated, $O(N)$ linear dialogue cleanup procedure that preserves system prompt limits by shrinking or omitting older turns' logs while leaving the last 4 turns raw. Implementation Hook: ContextManager::compact / server-rs/src/agent/context_manager.rs

Department Lead

[Tier 1: Operator] Type: Terminology Purpose: The primary coordinator agent in a workspace responsible for breaking down goals, recruiting specialist sub-agents, and compiling reports. Represents the management tier in the command chain. Implementation Hook: RoleAuthorityLevel::Management / state.registry.list_active_specialists()

Department Scope

[Tier 1: Operator] Type: Terminology Purpose: A communications constraint in the SovereignChat that restricts broadcasted instructions to all agents and sub-agents assigned to a specific workspace folder or cluster. Implementation Hook: SovereignChat / RunContext::department

Engine Shutdown

[Tier 3: Security] Type: Terminology Purpose: The graceful server termination protocol that persists all live agent states to the database and cleanly exits the backend process. Implementation Hook: POST /v1/engine/shutdown / routes::engine_control::shutdown_engine

Institutional Knowledge Store (IKS)

[Tier 1: Operator] Type: Terminology Purpose: The local, SQLite-backed repository managing swarm-wide insights, memory vectors, and validated operational procedures. Implementation Hook: ResourceHub::knowledge_store / routes::knowledge

Kill Switch

[Tier 3: Security] Type: Terminology Purpose: An emergency intervention mechanism that immediately halts all active agent thinking and execution loops and rejects pending oversight requests. Implementation Hook: POST /v1/engine/kill / routes::engine_control::kill_agents

Master Passphrase

[Tier 3: Security] Type: Terminology Purpose: The cryptographic password entered by the operator to derive key material via PBKDF2 for decrypting stored provider API keys in memory. Implementation Hook: crypto.worker.ts / use_vault_store

Mission Cluster

[Tier 1: Operator]

Note

Redirect: This term has been renamed to Project Workspace. See Project Workspace for the canonical definition.

Neural Vault

[Tier 3: Security]

Note

Redirect: This term has been renamed to Secure Credentials Vault. See Secure Credentials Vault for the canonical definition.

Open Knowledge Format (OKF)

[Tier 1: Operator] Type: Terminology Purpose: A standardized, self-describing draft specification (v0.1) for structuring institutional knowledge nodes with structured frontmatter metadata fields. Implementation Hook: KnowledgeEntry / AddKnowledgeRequest

Parity Guard

[Tier 2: Admin] Type: Tooling Purpose: A Python-driven validation command-line script that scans code files, routes, manifests, and documentation trees to flag drift or omissions. Implementation Hook: execution/parity_guard.py

Private Swarm Environment

[Tier 1: Operator] Type: Terminology Purpose: The secure, local-first execution environment where virtual agents run tools, process documents, and interact with LLMs without external data leaks. Implementation Hook: AppState / AgentRunner

Project Workspace

[Tier 1: Operator] Type: Terminology Purpose: An isolated local project directory containing state files, agent logs, configurations, and workspaces grouped toward a single business goal. Implementation Hook: GovernanceHub::max_clusters / RunContext::mission_id

RAG Scope

[Tier 2: Admin] Type: Terminology Purpose: The data ingestion context constraint limiting an agent's retrieval to specific files, workspace directories, or long-term memory. Implementation Hook: ResourceHub::symbol_graph / ResourceHub::code_graph

Secure Credentials Vault

[Tier 3: Security] Type: Terminology Purpose: Client-side credentials storage encrypted using PBKDF2 + AES-256-GCM via standard SubtleCrypto APIs. Stores LLM provider credentials securely in the browser. Implementation Hook: use_vault_store / crypto.worker.ts

Sovereign Reality

[Tier 1: Operator]

Note

Redirect: This term has been renamed to Private Swarm Environment. See Private Swarm Environment for the canonical definition.

STATIC_DIR

[Tier 2: Admin] Type: Environment Variable Purpose: The environment configuration specifying the directory path from which static frontend assets are served by the AXUM web server. Implementation Hook: std::env::var("STATIC_DIR") / server-rs/src/router.rs

Swarm Depth

[Tier 2: Admin] Type: Terminology Purpose: The recursion level of sub-agent recruitment within a mission. Enforced to prevent infinite delegation loops. Implementation Hook: MAX_SWARM_DEPTH / GovernanceHub::max_swarm_depth

Swarm Pulse

[Tier 3: Security] Type: Terminology Purpose: A real-time binary telemetry event stream showing system communications, agent thinking state, and tool calls. Implementation Hook: CommunicationHub::pulse_tx / SwarmPulse

---

🏗️ Compliance & Quality Standards

ALCOA+

[Tier 3: Security] Type: Standard Purpose: International data integrity guidelines (Attributable, Legible, Contemporaneous, Original, Accurate, Complete, Consistent, Enduring, Available) enforced across the engine telemetry streams. Implementation Hook: server-rs/src/security/audit.rs

GxP

[Tier 3: Security] Type: Standard Purpose: Good Practice quality guidelines and regulations (e.g., GCP, GMP, GLP) to ensure the system is safe for regulated domains. Implementation Hook: execution/verify_all.py

Software Validation Package (SVP)

[Tier 3: Security] Type: Compliance Artifact Purpose: Formally compiled validation documentation package (including FMEA risk assessment, OQ/PQ test scripts) for regulatory sign-off. Implementation Hook: docs/TEST_MISSIONS.md

---