Tauri IPC Commands Reference

Complete reference for all 58 Tauri IPC commands exposed by the Morphee Rust backend. These commands are invoked from the frontend via invoke() (Tauri IPC).

All commands that accept a group_id parameter validate it against the authenticated session (H-AUTHZ-006). Commands return Result<T, MorpheeError> where errors are serialized to the frontend as structured error objects.

---

Summary Table

#	Command	Category	Return Type	Auth
1	embed_text	Embeddings	EmbeddingResult	No
2	embed_batch	Embeddings	Vec<EmbeddingResult>	No
3	get_embedding_info	Embeddings	EmbeddingInfo	No
4	download_embedding_model	Embeddings	bool	No
5	is_model_cached	Embeddings	bool	No
6	memory_insert	Memory (Vector Store)	String	Yes
7	memory_search	Memory (Vector Store)	Vec<SearchResult>	Yes
8	memory_delete	Memory (Vector Store)	bool	Yes
9	memory_get	Memory (Vector Store)	Option<MemoryRecord>	Yes
10	memory_count	Memory (Vector Store)	usize	Yes
11	git_init_repo	Git Storage	String	Yes
12	git_save_conversation	Git Storage	String	Yes
13	git_save_memory	Git Storage	String	Yes
14	git_delete_memory	Git Storage	bool	Yes
15	git_sync	Git Storage	bool	Yes
16	git_create_branch	Git Branching	String	Yes
17	git_switch_branch	Git Branching	String	Yes
18	git_merge_branch	Git Branching	String	Yes
19	git_list_branches	Git Branching	Vec<String>	Yes
20	git_get_current_branch	Git Branching	String	Yes
21	git_get_commit_history	Git Branching	Vec<CommitInfo>	Yes
22	vault_get	Vault	Option<String>	Yes
23	vault_set	Vault	()	Yes
24	vault_delete	Vault	()	Yes
25	vault_exists	Vault	bool	Yes
26	fs_list_files	Filesystem	Vec<FileInfo>	Yes
27	fs_read_file	Filesystem	FileContent	Yes
28	fs_write_file	Filesystem	WriteResult	Yes
29	fs_delete_file	Filesystem	DeleteResult	Yes
30	fs_search_files	Filesystem	Vec<SearchMatch>	Yes
31	set_session	Session	()	No
32	clear_session	Session	()	No
33	health_check	Health	HealthStatus	No
34	queue_action	Offline Queue	String	Yes
35	drain_queue	Offline Queue	Vec<QueuedAction>	Yes
36	get_queue_status	Offline Queue	QueueStatus	Yes
37	remove_queued_action	Offline Queue	bool	Yes
38	clear_queue	Offline Queue	usize	Yes
39	morph_discover	OpenMorph	Vec<MorphDirInfo>	No
40	morph_init	OpenMorph	MorphInitResult	No
41	extension_load	WASM Extensions	String	No
42	extension_execute	WASM Extensions	ExtensionExecutionResult	No
43	extension_unload	WASM Extensions	bool	No
44	extension_list	WASM Extensions	Vec<String>	No
45	llm_chat_stream	Local LLM	String	Yes
46	llm_cancel_stream	Local LLM	()	Yes
47	llm_get_info	Local LLM	Option<LlmModelInfo>	No
48	llm_load_model	Local LLM	LlmModelInfo	Yes
49	llm_unload_model	Local LLM	()	Yes
50	llm_list_models	Local LLM	Vec<LlmModelInfo>	No
51	llm_download_model	Local LLM	()	Yes
52	llm_delete_model	Local LLM	()	Yes
53	vector_route_message	VectorRouter	VectorRouteResult	Yes
54	tts_speak	Audio: TTS	()	Yes
55	tts_stop	Audio: TTS	()	Yes
56	tts_is_speaking	Audio: TTS	bool	No
57	tts_set_rate	Audio: TTS	()	Yes
58	whisper_transcribe	Audio: Whisper STT	TranscriptionResult	Yes

---

1. Embeddings (5 commands)

Commands for generating text embeddings using ONNX (fastembed on desktop, candle on mobile).

embed_text

Generate an embedding vector for a single text string.

invoke('embed_text', { text: string }): Promise<EmbeddingResult>

Parameter	Type	Required	Description
text	String	Yes	Text to embed (max 8192 characters)

Returns: EmbeddingResult — { vector: number[], model: string, dimensions: number }

CPU-intensive ONNX inference runs on spawn_blocking to avoid starving the async runtime.

---

embed_batch

Generate embedding vectors for multiple texts in a single call.

invoke('embed_batch', { texts: string[] }): Promise<EmbeddingResult[]>

Parameter	Type	Required	Description
texts	Vec<String>	Yes	Array of texts to embed (max 100 items, each max 8192 chars)

Returns: Vec<EmbeddingResult> — one result per input text.

---

get_embedding_info

Get metadata about the loaded embedding model.

invoke('get_embedding_info'): Promise<EmbeddingInfo>

No parameters.

Returns: EmbeddingInfo — model name, dimensions (384 for AllMiniLML6V2), etc.

---

download_embedding_model

Download the embedding model for mobile (candle). No-op on desktop (fastembed auto-downloads).

invoke('download_embedding_model'): Promise<boolean>

No parameters.

Returns: bool — true if model is ready, false if not applicable (desktop).

---

is_model_cached

Check if the embedding model is cached locally.

invoke('is_model_cached'): Promise<boolean>

No parameters.

Returns: bool — always true on desktop (fastembed auto-downloads), checks cache on mobile.

---

2. Memory / Vector Store (5 commands)

LanceDB-backed vector memory with group-based isolation. Desktop uses LanceDB, mobile uses SQLite.

memory_insert

Insert a memory record. Automatically embeds the content before storing.

invoke('memory_insert', {
  content: string,
  memory_type: string,
  scope: string,
  group_id: string,
  space_id?: string,
  user_id?: string,
  source_conversation_id?: string,
  metadata?: object
}): Promise<string>

Parameter	Type	Required	Description
content	String	Yes	Text content to store and embed
memory_type	String	Yes	Type classification (e.g., "fact", "preference", "event", "skill_index")
scope	String	Yes	Visibility scope
group_id	String	Yes	Group ID (validated against session)
space_id	Option<String>	No	Space ID for space-scoped memories
user_id	Option<String>	No	User ID for user-scoped memories
source_conversation_id	Option<String>	No	Originating conversation ID
metadata	Option<Value>	No	Arbitrary JSON metadata

Returns: String — the generated memory record ID.

---

memory_search

Semantic search across memory records using cosine similarity.

invoke('memory_search', {
  query: string,
  group_id: string,
  scope?: string,
  space_id?: string,
  user_id?: string,
  memory_type?: string,
  limit?: number,
  threshold?: number
}): Promise<SearchResult[]>

Parameter	Type	Required	Description
query	String	Yes	Natural language search query (embedded for similarity)
group_id	String	Yes	Group ID (validated against session)
scope	Option<String>	No	Filter by scope
space_id	Option<String>	No	Filter by space
user_id	Option<String>	No	Filter by user
memory_type	Option<String>	No	Filter by memory type
limit	Option<usize>	No	Max results (default: 5)
threshold	Option<f32>	No	Min similarity score 0.0-1.0 (default: 0.3)

Returns: Vec<SearchResult> — matching records with similarity scores.

---

memory_delete

Delete a memory record by ID. Verifies the record belongs to the specified group before deleting.

invoke('memory_delete', { memory_id: string, group_id: string }): Promise<boolean>

Parameter	Type	Required	Description
memory_id	String	Yes	ID of the memory record to delete
group_id	String	Yes	Group ID (validated; record must belong to this group)

Returns: bool — true if deleted.

---

memory_get

Retrieve a single memory record by ID. Enforces group-based isolation.

invoke('memory_get', { memory_id: string, group_id: string }): Promise<MemoryRecord | null>

Parameter	Type	Required	Description
memory_id	String	Yes	ID of the memory record
group_id	String	Yes	Group ID (validated; record must belong to this group)

Returns: Option<MemoryRecord> — the record, or null if not found.

---

memory_count

Count total memory records for a group.

invoke('memory_count', { group_id: string }): Promise<number>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID (validated against session)

Returns: usize — total record count.

---

3. Git Storage (5 commands)

Git-backed persistent storage using libgit2. One repo per space. Stores conversations and memories as Markdown with YAML frontmatter.

git_init_repo

Initialize a git repository for a space.

invoke('git_init_repo', { group_id: string, space_id: string }): Promise<string>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID

Returns: String — path to the initialized repository.

---

git_save_conversation

Save a conversation to the git store as a Markdown file with YAML frontmatter.

invoke('git_save_conversation', {
  group_id: string,
  space_id: string,
  conversation_id: string,
  messages: MessageData[],
  title: string
}): Promise<string>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID
conversation_id	String	Yes	Conversation UUID
messages	Vec<MessageData>	Yes	Array of { role, content } messages
title	String	Yes	Conversation title

Returns: String — git commit hash.

---

git_save_memory

Save a memory entry to the git store.

invoke('git_save_memory', {
  group_id: string,
  space_id: string,
  memory_id: string,
  content: string,
  memory_type: string,
  metadata?: object
}): Promise<string>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID
memory_id	String	Yes	Memory record UUID
content	String	Yes	Memory content text
memory_type	String	Yes	Type classification
metadata	Option<Value>	No	Additional JSON metadata

Returns: String — git commit hash.

---

git_delete_memory

Delete a memory entry from the git store.

invoke('git_delete_memory', {
  group_id: string,
  space_id: string,
  memory_id: string,
  memory_type: string
}): Promise<boolean>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID
memory_id	String	Yes	Memory record UUID to delete
memory_type	String	Yes	Type of memory (determines file path)

Returns: bool — true if deleted.

---

git_sync

Sync a space's git repository with a remote URL (push/pull).

invoke('git_sync', {
  group_id: string,
  space_id: string,
  remote_url: string
}): Promise<boolean>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID
remote_url	String	Yes	Git remote URL to sync with

Returns: bool — true if sync succeeded.

---

4. Git Branching (6 commands)

Phase 3i.1: Git branching operations for temporal navigation and parallel space states.

git_create_branch

Create a new branch in a space's repository.

invoke('git_create_branch', {
  group_id: string,
  space_id: string,
  branch_name: string,
  from_branch?: string
}): Promise<string>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID
branch_name	String	Yes	Name for the new branch
from_branch	Option<String>	No	Source branch (defaults to current branch)

Returns: String — the created branch name.

---

git_switch_branch

Switch the space's repository to a different branch.

invoke('git_switch_branch', {
  group_id: string,
  space_id: string,
  branch_name: string
}): Promise<string>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID
branch_name	String	Yes	Branch to switch to

Returns: String — the active branch name after switch.

---

git_merge_branch

Merge a source branch into a target branch.

invoke('git_merge_branch', {
  group_id: string,
  space_id: string,
  source_branch: string,
  target_branch?: string,
  message?: string
}): Promise<string>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID
source_branch	String	Yes	Branch to merge from
target_branch	Option<String>	No	Branch to merge into (defaults to current)
message	Option<String>	No	Custom merge commit message

Returns: String — merge commit hash or result message.

---

git_list_branches

List all branches in a space's repository.

invoke('git_list_branches', {
  group_id: string,
  space_id: string
}): Promise<string[]>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID

Returns: Vec<String> — list of branch names.

---

git_get_current_branch

Get the name of the currently checked-out branch.

invoke('git_get_current_branch', {
  group_id: string,
  space_id: string
}): Promise<string>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID

Returns: String — current branch name (e.g., "main").

---

git_get_commit_history

Get the commit history for a branch.

invoke('git_get_commit_history', {
  group_id: string,
  space_id: string,
  branch?: string,
  limit: number
}): Promise<CommitInfo[]>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
space_id	String	Yes	Space ID
branch	Option<String>	No	Branch to read history from (defaults to current)
limit	usize	Yes	Max number of commits to return

Returns: Vec<CommitInfo> — commit objects with hash, message, author, timestamp.

---

5. Vault (4 commands)

Secure credential storage via OS keychain (macOS Keychain, Windows Credential Manager, Linux Secret Service). Android uses SQLite vault.

vault_get

Retrieve a secret from the vault.

invoke('vault_get', { key: string, group_id: string }): Promise<string | null>

Parameter	Type	Required	Description
key	String	Yes	Secret key name
group_id	String	Yes	Group ID (validated against session)

Returns: Option<String> — the secret value, or null if not found.

---

vault_set

Store a secret in the vault.

invoke('vault_set', { key: string, value: string, group_id: string }): Promise<void>

Parameter	Type	Required	Description
key	String	Yes	Secret key name
value	String	Yes	Secret value to store
group_id	String	Yes	Group ID (validated against session)

Returns: () (void)

---

vault_delete

Delete a secret from the vault.

invoke('vault_delete', { key: string, group_id: string }): Promise<void>

Parameter	Type	Required	Description
key	String	Yes	Secret key to delete
group_id	String	Yes	Group ID (validated against session)

Returns: () (void)

---

vault_exists

Check if a key exists in the vault.

invoke('vault_exists', { key: string, group_id: string }): Promise<boolean>

Parameter	Type	Required	Description
key	String	Yes	Secret key to check
group_id	String	Yes	Group ID (validated against session)

Returns: bool — true if the key exists.

---

6. Filesystem (5 commands)

Sandboxed file operations per group with path traversal prevention.

fs_list_files

List files and directories in a group's file storage area.

invoke('fs_list_files', { group_id: string, path?: string }): Promise<FileInfo[]>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
path	Option<String>	No	Relative subdirectory path (default: root)

Returns: Vec<FileInfo> — file metadata objects (name, size, type, modified date).

---

fs_read_file

Read a file from the group's storage.

invoke('fs_read_file', { group_id: string, path: string }): Promise<FileContent>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
path	String	Yes	Relative file path

Returns: FileContent — file content and metadata.

---

fs_write_file

Write content to a file in the group's storage.

invoke('fs_write_file', { group_id: string, path: string, content: string }): Promise<WriteResult>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
path	String	Yes	Relative file path
content	String	Yes	File content to write

Returns: WriteResult — write confirmation with path and bytes written.

---

fs_delete_file

Delete a file from the group's storage.

invoke('fs_delete_file', { group_id: string, path: string }): Promise<DeleteResult>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
path	String	Yes	Relative file path to delete

Returns: DeleteResult — deletion confirmation.

---

fs_search_files

Search file contents within a group's storage.

invoke('fs_search_files', {
  group_id: string,
  query: string,
  path?: string,
  max_results?: number
}): Promise<SearchMatch[]>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
query	String	Yes	Search query string
path	Option<String>	No	Subdirectory to search within (default: root)
max_results	Option<usize>	No	Max results (default: 20)

Returns: Vec<SearchMatch> — matching files with context.

---

7. Session (2 commands)

Session management for offline grace period support. The frontend calls these after authentication to persist the session across app restarts.

set_session

Set the authenticated session after login. Persists to vault for 72-hour offline grace period.

invoke('set_session', { group_id: string, expires_at?: number }): Promise<void>

Parameter	Type	Required	Description
group_id	String	Yes	Authenticated user's group ID (cannot be empty)
expires_at	Option<i64>	No	Token expiry as Unix timestamp (seconds)

Returns: () (void)

The session is stored both in-memory and in the vault (session/group_id, session/expires_at). On app restart, restore_session_from_vault (internal, not an IPC command) restores the session if within the 72-hour grace period.

---

clear_session

Clear the authenticated session on logout. Removes from both memory and vault.

invoke('clear_session'): Promise<void>

No parameters.

Returns: () (void)

---

8. Health (1 command)

health_check

Get the health status of all Tauri Rust subsystems.

invoke('health_check'): Promise<HealthStatus>

No parameters.

Returns: HealthStatus:

{
  healthy: boolean;        // Overall health (platform-aware criteria)
  platform: string;        // "macos" | "windows" | "linux" | "ios" | "android"
  subsystems: SubsystemStatus[];  // Per-subsystem status
  startup_errors: string[];       // Errors from initialization
}

Each SubsystemStatus has { name, ready, detail }. Checked subsystems:

• embeddings — ONNX/candle model
• vector_store — LanceDB/SQLite
• git_store — libgit2
• vault — OS keychain/SQLite
• action_queue — offline replay buffer

Health criteria by platform:

• Desktop: embeddings + git_store + vault must be ready
• Mobile with ML: embeddings + vector_store + git_store must be ready
• Mobile without ML: git_store must be ready

---

9. Offline Action Queue (5 commands)

Persistent queue for actions that fail while offline. Stored as JSON on disk. Actions are replayed when connectivity is restored.

queue_action

Add an action to the offline queue.

invoke('queue_action', {
  group_id: string,
  action_type: string,
  payload: object
}): Promise<string>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID
action_type	String	Yes	Action type identifier (e.g., "send_message", "create_task")
payload	Value	Yes	Arbitrary JSON payload for replay

Returns: String — the generated action ID.

---

drain_queue

Drain all pending actions for a group (oldest-first). Increments each action's attempts counter. Call remove_queued_action after successfully replaying each action.

invoke('drain_queue', { group_id: string }): Promise<QueuedAction[]>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID

Returns: Vec<QueuedAction> — all pending actions for the group, ordered oldest-first.

---

get_queue_status

Get queue status without modifying the queue.

invoke('get_queue_status', { group_id: string }): Promise<QueueStatus>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID

Returns: QueueStatus — pending count and oldest action timestamp.

---

remove_queued_action

Remove a successfully replayed action from the queue.

invoke('remove_queued_action', {
  action_id: string,
  group_id: string
}): Promise<boolean>

Parameter	Type	Required	Description
action_id	String	Yes	ID of the action to remove
group_id	String	Yes	Group ID

Returns: bool — true if the action was found and removed.

---

clear_queue

Discard all pending actions for a group.

invoke('clear_queue', { group_id: string }): Promise<number>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID

Returns: usize — number of actions cleared.

---

10. OpenMorph (2 commands)

V1.0: Local .morph/ directory discovery and initialization for git-native Spaces.

morph_discover

Scan a local directory tree for existing Morph Spaces (directories containing .morph/). This is a pre-authentication operation -- no group_id required.

invoke('morph_discover', {
  root_path?: string,
  max_depth?: number
}): Promise<MorphDirInfo[]>

Parameter	Type	Required	Description
root_path	Option<String>	No	Directory to start scanning (default: user home). Must be >2 chars, cannot be filesystem root.
max_depth	Option<u32>	No	Max directory depth (default: 5, capped at 10)

Returns: Vec<MorphDirInfo> — discovered .morph/ directories with metadata.

---

morph_init

Initialize a .morph/ Morph Space in an arbitrary local directory. Idempotent: returns existing info if .morph/ already exists.

Creates:

• .morph/config.yaml -- space identity and sync configuration
• .morph/canvas.yaml -- blank canvas state
• .morph/acl.yaml -- default access control
• .morph/.git/ -- git repository (non-fatal if git init fails)

invoke('morph_init', {
  path: string,
  space_name: string,
  space_id: string,
  sync_mode?: string,
  remote_url?: string
}): Promise<MorphInitResult>

Parameter	Type	Required	Description
path	String	Yes	Absolute path to the directory
space_name	String	Yes	Human-readable space name
space_id	String	Yes	UUID assigned by the backend
sync_mode	Option<String>	No	"local-only" (default), "morphee-hosted", or "git-remote"
remote_url	Option<String>	No	Git remote URL for git-remote sync mode

Returns: MorphInitResult -- initialization result with space metadata.

---

11. WASM Extensions (4 commands)

V1.2: Load and execute WASM extensions in a sandboxed runtime.

extension_load

Load a WASM extension from raw bytes with a manifest.

invoke('extension_load', {
  wasm_bytes: number[],   // Uint8Array serialized
  manifest_json: string   // JSON string of ExtensionManifest
}): Promise<string>

Parameter	Type	Required	Description
wasm_bytes	Vec<u8>	Yes	Raw WASM binary bytes
manifest_json	String	Yes	JSON-serialized ExtensionManifest

Returns: String — the extension ID.

---

extension_execute

Execute an action on a loaded extension.

invoke('extension_execute', {
  extension_id: string,
  action: string,
  params: object
}): Promise<ExtensionExecutionResult>

Parameter	Type	Required	Description
extension_id	String	Yes	ID of the loaded extension
action	String	Yes	Action name to execute
params	Value	Yes	JSON parameters to pass to the action

Returns: ExtensionExecutionResult — execution result with output data.

---

extension_unload

Unload a WASM extension from the runtime.

invoke('extension_unload', { extension_id: string }): Promise<boolean>

Parameter	Type	Required	Description
extension_id	String	Yes	ID of the extension to unload

Returns: bool — true if the extension was found and unloaded.

---

extension_list

List all currently loaded extension IDs.

invoke('extension_list'): Promise<string[]>

No parameters.

Returns: Vec<String> — list of loaded extension IDs.

---

12. Local LLM (8 commands)

V1.5: Local GGUF model inference via candle. Supports Phi-4 Mini (3.8B) and Llama 3.2 3B. Uses Metal acceleration on macOS.

llm_chat_stream

Start streaming LLM token generation. Returns immediately with a generation_id. Tokens are emitted as Tauri events.

invoke('llm_chat_stream', {
  request: {
    generation_id: string,
    messages: Array<{ role: string, content: string }>,
    group_id: string,
    config?: {
      max_tokens: number,       // default: 1024
      temperature: number,      // default: 0.7
      top_p: number,            // default: 0.9
      repeat_penalty: number,   // default: 1.1
      repeat_last_n: number     // default: 64
    }
  }
}): Promise<string>

Parameter	Type	Required	Description
request	ChatStreamRequest	Yes	Chat request with messages and config

Returns: String — the generation_id (echoed back).

Emitted events:

• llm-token — { generation_id, token } for each generated token (real-time streaming)
• llm-done — { generation_id, token_count } when generation completes
• llm-error — { generation_id, message } on error

Example usage:

import { invoke } from '@tauri-apps/api/core';
import { listen } from '@tauri-apps/api/event';

const unlisten = await listen('llm-token', (event) => {
  appendToUI(event.payload.token);
});

await invoke('llm_chat_stream', {
  request: {
    generation_id: crypto.randomUUID(),
    messages: [
      { role: 'system', content: 'You are a helpful assistant.' },
      { role: 'user', content: 'Hello!' }
    ],
    group_id: currentGroupId
  }
});

---

llm_cancel_stream

Cancel an in-progress generation by setting the atomic cancel flag.

invoke('llm_cancel_stream', { group_id: string }): Promise<void>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID (validated against session)

Returns: () (void). The running generation will stop at the next token boundary.

---

llm_get_info

Get info about the currently loaded model.

invoke('llm_get_info'): Promise<LlmModelInfo | null>

No parameters.

Returns: Option<LlmModelInfo>:

{
  model_id: string;        // e.g., "phi-4-mini-q4"
  display_name: string;    // e.g., "Phi-4 Mini (Q4)"
  parameters: string;      // e.g., "3.8B"
  quantization: string;    // e.g., "Q4_K_M"
  context_length: number;  // e.g., 4096
  downloaded: boolean;
  size_bytes: number | null;
}

Returns null if no model is loaded.

---

llm_load_model

Load a previously downloaded model into memory. Emits progress events during loading (5-30 seconds).

invoke('llm_load_model', {
  model_id: string,
  group_id: string
}): Promise<LlmModelInfo>

Parameter	Type	Required	Description
model_id	String	Yes	Model identifier: "phi-4-mini-q4" or "llama-3.2-3b-q4"
group_id	String	Yes	Group ID

Returns: LlmModelInfo — info about the loaded model.

Emitted events:

• llm-load-progress — { stage: "tokenizer" | "weights" | "ready", progress: 0.0-1.0 }

Errors if the model has not been downloaded first.

---

llm_unload_model

Unload the current model from memory to free resources.

invoke('llm_unload_model', { group_id: string }): Promise<void>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID

Returns: () (void)

---

llm_list_models

List all available models in the catalog with their download status.

invoke('llm_list_models'): Promise<LlmModelInfo[]>

No parameters.

Returns: Vec<LlmModelInfo> — all models in the catalog.

Available models:

Model ID	Display Name	Parameters	Context	HuggingFace Repo
phi-4-mini-q4	Phi-4 Mini (Q4)	3.8B	4096	microsoft/Phi-4-mini-instruct-GGUF
llama-3.2-3b-q4	Llama 3.2 3B (Q4)	3.2B	4096	bartowski/Llama-3.2-3B-Instruct-GGUF

---

llm_download_model

Download a model from HuggingFace. Emits progress events during download. No-op if already downloaded.

invoke('llm_download_model', {
  model_id: string,
  group_id: string
}): Promise<void>

Parameter	Type	Required	Description
model_id	String	Yes	Model identifier to download
group_id	String	Yes	Group ID

Returns: () (void)

Emitted events:

• llm-download-progress — { model_id, bytes_downloaded, bytes_total }

---

llm_delete_model

Delete a downloaded model from disk. Automatically unloads the model if it is currently active.

invoke('llm_delete_model', {
  model_id: string,
  group_id: string
}): Promise<void>

Parameter	Type	Required	Description
model_id	String	Yes	Model identifier to delete
group_id	String	Yes	Group ID

Returns: () (void)

---

13. VectorRouter (1 command)

V1.5: Offline-first message routing. Checks local LanceDB memory and skill index before falling back to LLM.

vector_route_message

Route a user message using local vector memory and skill index. Determines whether the query can be answered directly from memory, routed to a skill, or requires the LLM.

invoke('vector_route_message', {
  message: string,
  group_id: string
}): Promise<VectorRouteResult>

Parameter	Type	Required	Description
message	String	Yes	User's message to route
group_id	String	Yes	Group ID

Returns: VectorRouteResult:

{
  type: "DIRECT_MEMORY" | "SKILL_EXECUTE" | "SKILL_HINT" | "LLM_REQUIRED";
  content?: string;     // For DIRECT_MEMORY: the stored answer text
  skill_id?: string;    // For SKILL_*: the skill UUID
  skill_name?: string;  // For SKILL_*: human-readable skill name
  score: number;        // Similarity score (0.0-1.0)
}

Routing thresholds (from VectorRouter):

• DIRECT_MEMORY >= 0.92 -- use content directly as AI response
• SKILL_EXECUTE >= 0.88 -- execute the matched skill (no required params)
• SKILL_HINT >= 0.83 -- pass skill name as hint in LLM system prompt
• LLM_REQUIRED -- no match above threshold; fall through to LLM

Falls back to LLM_REQUIRED gracefully if embedding provider or vector store are not initialized.

---

14. Audio: TTS (4 commands)

V1.5: Text-to-speech via platform TTS engines (macOS AVSpeech, Windows SAPI5, Linux espeak/piper). Requires --features audio Cargo flag.

tts_speak

Speak text using the platform TTS engine. Initializes the TTS provider lazily on first call. Non-blocking -- speech continues in background.

invoke('tts_speak', { text: string, group_id: string }): Promise<void>

Parameter	Type	Required	Description
text	String	Yes	Text to speak (max 4096 characters)
group_id	String	Yes	Group ID

Returns: () (void)

---

tts_stop

Stop any currently-in-progress TTS speech.

invoke('tts_stop', { group_id: string }): Promise<void>

Parameter	Type	Required	Description
group_id	String	Yes	Group ID

Returns: () (void). No-op if nothing is playing.

---

tts_is_speaking

Check if TTS is currently speaking.

invoke('tts_is_speaking'): Promise<boolean>

No parameters.

Returns: bool — true if audio is currently playing.

---

tts_set_rate

Set TTS speech rate.

invoke('tts_set_rate', { rate: number, group_id: string }): Promise<void>

Parameter	Type	Required	Description
rate	f32	Yes	Speech rate: 0.5 = slow, 1.0 = normal, 2.0 = fast (clamped to 0.1-3.0)
group_id	String	Yes	Group ID

Returns: () (void)

---

15. Audio: Whisper STT (1 command)

V1.5: Local speech-to-text via Whisper-tiny ONNX model. Status: skeleton with API surface; ONNX inference pipeline not yet implemented.

whisper_transcribe

Transcribe PCM audio samples to text using the local Whisper model.

invoke('whisper_transcribe', {
  audio_data: number[],  // Float32Array serialized as Vec<f32>
  sample_rate: number,
  group_id: string
}): Promise<TranscriptionResult>

Parameter	Type	Required	Description
audio_data	Vec<f32>	Yes	Mono PCM f32 samples (max 30s at 16kHz = 480,000 samples)
sample_rate	u32	Yes	Audio sample rate (should be 16000 for Whisper)
group_id	String	Yes	Group ID

Returns: TranscriptionResult:

{
  text: string;             // Transcribed text
  confidence: number | null; // Confidence score [0.0, 1.0] if available
  duration_ms: number;       // Processing time in milliseconds
}

Falls back to error if the Whisper model has not been downloaded. The frontend should fall back to the Web Speech API.

---

Tauri Event Reference

Several commands emit events via app.emit() rather than returning data directly. Listen for these on the frontend using @tauri-apps/api/event.

Event Name	Payload	Emitted By
llm-token	{ generation_id: string, token: string }	llm_chat_stream
llm-done	{ generation_id: string, token_count: number }	llm_chat_stream
llm-error	{ generation_id: string, message: string }	llm_chat_stream
llm-download-progress	{ model_id: string, bytes_downloaded: number, bytes_total: number }	llm_download_model
llm-load-progress	{ stage: string, progress: number }	llm_load_model

---

Error Handling

All commands return Result<T, MorpheeError>. The MorpheeError enum maps to specific subsystem errors:

Variant	Description
MorpheeError::Embedding(String)	Embedding provider errors
MorpheeError::VectorStore(String)	LanceDB/SQLite vector store errors
MorpheeError::Git(String)	Git storage errors
MorpheeError::Vault(String)	Keychain/vault errors
MorpheeError::Filesystem(String)	File store errors
MorpheeError::Auth(String)	Session validation failures
MorpheeError::Extension(String)	WASM extension errors
MorpheeError::Llm(String)	Local LLM, TTS, and Whisper errors

On the frontend, invoke errors are caught as rejected promises with the error message string.

---

Feature Flags

Some commands require specific Cargo features to be functional:

Feature	Commands Affected	Description
local-llm	llm_chat_stream, llm_load_model, llm_cancel_stream	Enables candle GGUF inference (desktop)
mobile-ml	download_embedding_model, is_model_cached	Enables candle embeddings + SQLite vector store (mobile)
audio	tts_speak, tts_stop, tts_is_speaking, tts_set_rate	Enables platform TTS via tts crate

Without the relevant feature flag, stub implementations return descriptive errors (e.g., "Build with --features local-llm").

---

Source Files

File	Commands
frontend/src-tauri/src/commands/embedding_commands.rs	embed_text, embed_batch, get_embedding_info, download_embedding_model, is_model_cached
frontend/src-tauri/src/commands/memory_commands.rs	memory_insert, memory_search, memory_delete, memory_get, memory_count
frontend/src-tauri/src/commands/git_commands.rs	git_init_repo, git_save_conversation, git_save_memory, git_delete_memory, git_sync, git_create_branch, git_switch_branch, git_merge_branch, git_list_branches, git_get_current_branch, git_get_commit_history
frontend/src-tauri/src/commands/vault_commands.rs	vault_get, vault_set, vault_delete, vault_exists
frontend/src-tauri/src/commands/fs_commands.rs	fs_list_files, fs_read_file, fs_write_file, fs_delete_file, fs_search_files
frontend/src-tauri/src/commands/session_commands.rs	set_session, clear_session
frontend/src-tauri/src/commands/health_commands.rs	health_check
frontend/src-tauri/src/commands/queue_commands.rs	queue_action, drain_queue, get_queue_status, remove_queued_action, clear_queue
frontend/src-tauri/src/commands/morph_commands.rs	morph_discover, morph_init
frontend/src-tauri/src/commands/extension_commands.rs	extension_load, extension_execute, extension_unload, extension_list
frontend/src-tauri/src/commands/llm_commands.rs	llm_chat_stream, llm_cancel_stream, llm_get_info, llm_load_model, llm_unload_model, llm_list_models, llm_download_model, llm_delete_model
frontend/src-tauri/src/commands/vector_commands.rs	vector_route_message
frontend/src-tauri/src/commands/audio_commands.rs	tts_speak, tts_stop, tts_is_speaking, tts_set_rate, whisper_transcribe
frontend/src-tauri/src/lib.rs	Command registration (invoke_handler)