Skip to main content

galet-tools

Self-contained generic tool handlers and the handler framework they run on.

Lineage

This package is extracted from the Lucy monorepo (src/handlers) as described in the Lucy design doc software/ai/lucy/design/handlers-extraction.md and the plan software/ai/lucy/plan/handlers-extraction.md.

It is a fresh repository with no shared git history. Handler logic moves verbatim; the only intentional changes are the import namespace (src.handlers.* -> galet_tools.*) and the configuration boundary: tools depend on narrow port interfaces owned by this package (ConfigProvider, StorageLocationResolver, SandboxRootResolver, SecurityPolicy, ...), never on Lucy's ConfigManager. Lucy implements those ports at its composition root.

Features

  • Handler framework — HandlerV2 ABC, HandlerRegistry, schema handling, and permission filtering.
  • Generic tools — file_load, file_save, patch_apply, execute_command, execute_command2, and generate_svg, registered under their existing names. file_load supports bounded line-range reads.
  • Repository navigation — repo_index and repo_search preserve Python AST navigation and also search supported JavaScript, TypeScript, C#, Java, and other source files using bounded text indexing.
  • Ports — sibling-owned interfaces the host implements, keeping the tools free of any host dependency.
  • Handler discovery — installed packages can contribute handlers through Python package entry points without changing Lucy or galet-tools.

Quick start

python -m venv .venv
.venv/bin/pip install -e .

Update the project and all dependencies

From the galet-tools directory, with your virtual environment active, run:

python -m pip install --upgrade --upgrade-strategy eager -e '.[memory,mcp,video]'

This installs your local galet-tools in editable mode and upgrades its runtime dependencies, including galet-memory and galet, to the newest versions allowed by pyproject.toml. It includes the memory, MCP, and video extras; exact version pins remain in effect.

Run the suite:

python -m pytest -q

Using the registry in a host

Galet owns the handler contract and base registry. A host such as Lucy composes the generic handlers, its own host-specific handlers, and installed extension packages:

from galet_tools import HandlerRegistry, register_installed_handlers
from galet_tools.register_generic_defaults import register_generic_defaults

registry = HandlerRegistry()
register_generic_defaults(registry)
register_lucy_handlers(registry)
register_installed_handlers(registry)

Discovery adds installed capabilities to the registry; it does not grant them to an agent. The host should continue to apply its agent and context allowlists when selecting tool definitions.

Repository navigation

repo_index maps supported source files, looks up a file or directory, and finds symbols. Python files use AST-based classes, functions, methods, and exact line ranges. Other supported languages, including JavaScript, TypeScript, and C#, use a lightweight text index that returns matching lines and line numbers. repo_search ranks both Python and other source files using paths and content; Python also uses structural metadata. Common dependency, generated, and build directories are skipped, and tests are omitted unless include_tests is true.

Creating a handler extension

An extension package declares a registrar in its pyproject.toml:

[project.entry-points."galet_tools.handlers"]
web = "galet_web_tools:register"

The target is a callable that receives the shared registry:

def register(registry):
    registry.register(SearchHandler)
    registry.register(FetchHandler)

Entry points are loaded in deterministic name order. A plugin that cannot be loaded, does not resolve to a callable, or fails during registration raises HandlerPluginError with the responsible entry-point name. This is intentionally fail-fast so a host cannot start with an unexpectedly incomplete tool set.

Named roots and relative paths

File, patch, repository, and command tools use one root selector: external_root, with an exact configured name. Model-facing schemas no longer expose location. Root names are arbitrary, case-sensitive aliases supplied by the host's effective configuration. For example:

{"external_root": "lucy_data_files", "path": "data/accounts/john/notes.txt"}

The resolver maps only the name to its configured directory. It does not infer special behavior from aliases such as sandbox, storage, or lucy_data_files, and it does not append account or namespace segments. Callers include those segments explicitly when needed. In Lucy, the host supplies the effective mapping after global/local configuration overrides; galet-tools does not load Lucy's configuration itself.

path (file/patch), root (repository), and working_directory (command) stay relative to the selected directory. . selects the directory itself for repository/command tools. Absolute paths and escapes in these handler fields are rejected. Command scripts and program arguments retain their own path requirements and may contain absolute paths; setting a working directory does not confine everything a program can access.

Compatibility: direct runtime calls with an explicit legacy location retain their storage/external or sandbox/external semantics, including legacy sandbox account scoping. Older file/repository/command calls that omit both selectors retain their historical defaults. A named root without location always uses the configured mapping, never a storage namespace or account sandbox. Legacy resolver methods and result location metadata remain for existing consumers. New model-facing calls require a non-empty external_root; configure roots explicitly rather than relying on legacy defaults.

Applying a single-file patch

patch_apply wraps git apply and requires Git on PATH, but the selected root does not need to be a Git repository. path is relative to that root: with external_root="repos", a file in a repository may have path="image_description/src/image_description/image/prompts.py". Keep that full root-relative path throughout validation and application.

Supply a numbered unified diff. Both Git-style a/ and b/ header prefixes and prefix-free headers are accepted. For path="repo/file.txt", for example:

--- a/repo/file.txt
+++ b/repo/file.txt
@@ -1 +1 @@
-old
+new

Alternatively use --- repo/file.txt and +++ repo/file.txt. The handler selects -p1 or -p0 only when Git reports exactly the requested file, then uses that same setting for checking and applying. It does not retry another setting after a content-check failure. Results report strip_components; path inspection failures include path_attempts and Git diagnostics.

Use check_only=true first to validate without writing, then submit the same patch with check_only=false to apply it. *** Begin Patch syntax is not supported. Wrong-file, multi-file, binary, and escaping-path patches remain rejected. Correct malformed hunk counts or headers rather than shortening path.

Memory tools example

The episodic_memory handler takes an injected episodic store, and curate_chat takes an injected curation service, and reset_session takes the same curation service. semantic_memory takes an injected galet_memory.semantic.SemanticMemory. procedural_memory exposes context/skill recall and scoped context/skill edits. The episodic and procedural tools are registered in the generic registry, with backends supplied by the host; MCP constructs them when explicitly allowlisted. Curation/reset/semantic tools still require application services and are not exposed by the generic MCP server.

See memory tools for procedural operations, correlation invalidation, configuration pointing at Lucy storage, and Inspector/smoke tests.

Run a small SQLite session through creation, event append and read, digest preview, archive, an active-session read, and an idempotent archive retry, and a context reset that retains history and session metadata. It then writes and indexes the archived digest and recalls it through VectorSemanticMemory in the digests namespace, using deterministic embedding and index ports without an API key:

python -m pip install -e '.[memory]'
python examples/curate_chat.py

The memory extra pins galet-memory PR #33’s merge commit until the coordinated release. This is a breaking episodic API/storage change; use fresh storage. Lucy and galet-prompt-builder still require consumer updates before deployment.

The example uses a deterministic digest generator and runs all four tool adapters with one temporary session. The example explicitly writes the digest text, embeds it, and inserts a StoredEmbedding into an in-memory index; curation does not automatically publish to semantic memory. It checks that recall is empty before publication and returns the digest afterward. In an application, inject galet_memory.curation.CurationService with the application's digest generator and construct CurateChatHandler(service) through a registry factory.

Video generation

video_generate accepts an uploaded image_id, a motion prompt, and duration_seconds (4, 6, or 8). These three arguments are required. The following arguments can be supplied for each request or omitted to use defaults:

Argument Default Supported values
source google Currently Google only; other providers fail explicitly
model veo-3.1-fast-generate-preview Google video model ID
aspect_ratio 9:16 9:16, 16:9
resolution 720p 720p, 1080p, 4k; the latter two require 8 seconds
preserve_subject true Boolean; adds identity/garment preservation instructions
add_sound false Boolean; false removes audio, true retains provider audio

Example tool arguments:

{
  "image_id": "<uploaded image UUID>",
  "prompt": "One gentle step forward and a slight turn.",
  "duration_seconds": 4,
  "source": "google",
  "model": "veo-3.1-fast-generate-preview",
  "aspect_ratio": "9:16",
  "resolution": "720p",
  "preserve_subject": true,
  "add_sound": false
}

Install galet-tools[video] and Galet's Gemini SDK dependency (galet[gemini]) for Google generation. Silent output also requires FFmpeg on the host PATH (e.g. sudo apt install ffmpeg on Debian/Ubuntu). The handler checks this before submitting a paid generation request. Google's Gemini Developer API does not accept the SDK's generate_audio setting, so silent output is implemented by removing all audio tracks after download, copying the video stream without re-encoding. This does not reduce provider generation charges. add_sound=true retains whatever audio the provider returns; it does not synthesize a separate soundtrack. Subject preservation is a prompt instruction, not a guarantee.

Results include the selected source, model, aspect ratio, resolution, preserve_subject, and add_sound alongside the video metadata. Defaults are applied by the handler, including for older callers supplying only the three original arguments. The tool schema uses strict=false so defaulted arguments can genuinely be omitted.

MCP server

An optional MCP (Model Context Protocol) server exposes the generic tools over streamable HTTP. Install the extra:

.venv/bin/pip install -e '.[mcp]'

Copy the example config and start the server:

cp galet-mcp.example.json galet-mcp.json
galet-tools-mcp

galet-tools-mcp reads galet-mcp.json from the current directory by default; point it elsewhere with --config <path>.

Configuration

galet-mcp.json mirrors galet-mcp.example.json. Required keys are account, storage_root, storage_namespace, and sandbox_root; a null, blank, or (for the two path keys) relative value fails validation. Optional keys left out or set to null fall back to their defaults:

  • allowlist — tools to expose; the default includes file_load, file_save, patch_apply, legacy execute_command, structured execute_command2, generate_svg.
  • external_roots — arbitrary named roots for file, patch, repository, and command tools; default none. Configure these for model-facing root-based calls. The example includes data-store and sandbox aliases, which have no special meaning to the resolver.
  • host — default 127.0.0.1; loopback only.
  • port — default 8765.
  • fonts — font paths for the generators; default none.
  • episodic_memory_db_path — optional absolute relational SQLite path; defaults to <storage_root>/<storage_namespace>/chat2.sqlite when episodic tools are allowed.
  • procedural_memory — optional root/layout/context-resolution overrides. See memory configuration.

What it exposes

tools/list and tools/call over streamable HTTP. With the default port, the endpoint is http://127.0.0.1:8765/mcp.

Test interactively with MCP Inspector

MCP Inspector provides a browser interface for testing galet-tools as an MCP user. Start galet-tools-mcp, then run:

npx --yes @modelcontextprotocol/inspector@latest --web --server-url http://127.0.0.1:8700/mcp --transport http

This command requires Node.js and npm. The server URL must use the port from your galet-mcp.json; the example above uses 8700. In the Inspector, use Tools to list the tools, inspect their schemas, enter arguments, and run real tools/call requests against the server.

Smoke-test from the CLI

galet-tools-smoke is the end-to-end CLI client: it connects over HTTP (streamable HTTP) to a running galet MCP server using the official MCP SDK and performs real tools/list and tools/call round trips.

galet-tools-smoke list
galet-tools-smoke list --json
galet-tools-smoke call <tool> '<json args>'

The default URL is http://127.0.0.1:8765/mcp; override it with --url. The default read timeout is 30s; override it with --timeout.

Example round trip (storage root from galet-mcp.json):

galet-tools-smoke call file_save '{"external_root":"lucy_data_files","path":"galet/smoke/hello.txt","file_content":"hello galet smoke","overwrite":true}'
galet-tools-smoke call file_load '{"external_root":"lucy_data_files","path":"galet/smoke/hello.txt","start_line":null,"line_count":null,"include_line_numbers":false}'
galet-tools-smoke call execute_command2 '{"mode":"process","executable":"python","arguments":["-c","print(123)"],"script":"","shell":"none","external_root":"sandbox","working_directory":".","timeout_seconds":30,"success_exit_codes":[0]}'

Exit codes: 0 ok; 1 tool reported an error (isError); 2 usage or bad JSON; 3 could not reach the server.

Fail-closed behaviour

The server refuses to start when the allowlist is empty or has no overlap with the registered tools, and only the advertised tools are callable. Unknown tool names in the allowlist fail validation.

Security

Binding is loopback-only by design: host must resolve to a loopback address. Remote access needs a separate reverse-proxy design.

License

MIT — see LICENSE.

Metadata

Release files for galet-tools 0.1.27

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for galet-tools 0.1.27
File Size Uploaded Provenance Actions
galet_tools-0.1.27.tar.gz 102.4 kB Without provenance Details for galet_tools-0.1.27.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for galet-tools 0.1.27
File Interpreter ABI Platform Actions
galet_tools-0.1.27-py3-none-any.whl Python 3 none any Details for galet_tools-0.1.27-py3-none-any.whl

Total release size: 184.0 kB

Release files / galet_tools-0.1.27.tar.gz

Download URL galet_tools-0.1.27.tar.gz
Size 102.4 kB
Tags Source
SHA-256 checksum
How to use checksums
e2231d4cd49708b53361ca0b8f3a4b1ec7a91a0feee8e1e2014d195754c22893
BLAKE2b-256 checksum
How to use checksums
66e4ac075302cc3543dbe4cdc45372ce463233bf04df048fab05da323a8601b0
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.12.16 {"installer":{"name":"uv","version":"0.12.16","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Debian GNU/Linux","version":"12","id":"bookworm","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Release files / galet_tools-0.1.27-py3-none-any.whl

Download URL galet_tools-0.1.27-py3-none-any.whl
Size 81.6 kB
Tags Python 3
SHA-256 checksum
How to use checksums
7d757be773b7aacdd895a1c3df5ab9bd8cd5c78b513d7fb98bcdc531625b7338
BLAKE2b-256 checksum
How to use checksums
ed92675f1aa7815a6536ec5c7b0a1dc18be6493f0383cdcdbd9f1ff206d40c00
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.12.16 {"installer":{"name":"uv","version":"0.12.16","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Debian GNU/Linux","version":"12","id":"bookworm","libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

Security for galet-tools 0.1.27

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.1.27 This release

2 release files No provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page