This PR introduces ClickHouse support by finishing off
https://github.com/googleapis/genai-toolbox/pull/713. We have had quite
a few requests internally and externally of people who want ClickHouse
support and wish to contribute but can't use the toolbox because of this
stalled PR. We also tried to reach out to @sidpan1 ourselves with no
success so we are going to assume the contributor doesn't have time to
complete. We would like to thank @sidpan1 for getting this going -
community contributions help make ClickHouse better for everyone.
This PR introduces
- Adds ClickHouse HTTPS and HTTP source support (drops native support
and compression in connection)
- Adds tools: `execute_sql`, ~~`describe_table`, `list_databases`,
`list_tables`~~
- Adds unit and integration tests
- Adds docs
Other contributors can build on this foundation
### Connect to local ClickHouse
```sh
CLICKHOUSE_HOST="localhost" \
CLICKHOUSE_PORT=8123 \
CLICKHOUSE_USER="default" \
CLICKHOUSE_PASSWORD="" \
CLICKHOUSE_DATABASE=default \
CLICKHOUSE_PROTOCOL="http" \
go run main.go --prebuilt clickhouse --ui --port 8080
```
### Connect to ClickHouse Cloud
```sh
CLICKHOUSE_HOST="tsmtweovmw.us-east-2.aws.clickhouse.cloud" \
CLICKHOUSE_PORT=8443 \
CLICKHOUSE_USER="default" \
CLICKHOUSE_PASSWORD="[REDACTED]" \
CLICKHOUSE_DATABASE=default \
CLICKHOUSE_PROTOCOL=https \
go run main.go --prebuilt clickhouse --ui --port 8080
```
### Run tests
```bash
go test -v ./tests/clickhouse/ -run TestClickHouse
go test -v ./tests/clickhouse/ -run TestClickHouseBasicConnection
```
<img width="1318" height="895" alt="Screenshot 2025-08-06 at 10 01 01"
src="https://github.com/user-attachments/assets/034d8f1b-10d6-4097-8033-5b0da93ad3fc"
/>
---------
Co-authored-by: Pete Hampton <pjhampton@users.noreply.github.com>
Co-authored-by: Wenxin Du <117315983+duwenxin99@users.noreply.github.com>
Co-authored-by: duwenxin <duwenxin@google.com>
Existing `/mcp` endpoint of Toolbox does not support auth (authorized
invocation and authenticated parameters). This PR add support for
Toolbox auth to the `/mcp` endpoint.
Added integration test for MCP with auth.
Note that Toolbox auth is **NOT** supported in stdio transport protocol,
invocations of tools with auth will result in error.
1. Add ask-data-insights tool based on conversational analytic API.
2. Add tokenSource for ask-data-insights tool, it uses access token
instead of client or restService.
3. Add a max row count to source, currently fixed to 50 and used only
for ask-data-insights tool. Later we may make it available for user to
make change and apply to bigquery-execute-sql and bigquery-sql to avoid
return too many data by accident.
---------
Co-authored-by: Yuan Teoh <45984206+Yuan325@users.noreply.github.com>
Co-authored-by: Averi Kitsch <akitsch@google.com>
For Toolbox protocol:
Before - return 400 error for all tool invocation errors.
After - Propagate auth-related errors (401 & 403) to the client if using
client credentials. If using ADC, raise 500 error instead.
For MCP protocol:
Before - return 200 with error message in the response body.
After - Propagate auth-related errors (401 & 403) to the client if using
client credentials. If using ADC, raise 500 error instead.
* fix not null expression
* exclude dimensions that end in _raw
* Add tags to return from get_dimensions, get_measures, etc.
* Add synonyms to return from get_dimensions, get_measures, etc.
* Instruct Agent to only create pie chart visualizations with one
measure and one dimension.
* Added docs about the return types of
get_dimensions/get_measures/get_filters/get_parameters.
---------
Co-authored-by: Yuan Teoh <45984206+Yuan325@users.noreply.github.com>
## Description
This Pull Request introduces support for the Firebird database as a new
`source` and adds the `firebird-sql` and `firebird-execute-sql` tools.
The implementation follows the architectural pattern established by the
existing PostgreSQL and MySQL tools.
Comprehensive integration tests have been added in the `tests/firebird`
directory, covering table creation, data insertion, and the invocation
of all related tools. Corresponding user and developer documentation has
also been included.
## Test Results
The core functionality is working as demonstrated by the vast majority
of tests passing. However, a few tests fail due to limitations and
inflexibilities in the generic test harness (`tests/tool.go`), which is
not fully compatible with Firebird's specific syntax and behavior.
### Known and Justified Failures:
* **`invoke_my-tool` / `invoke_my-array-tool`:** The `RunToolInvokeTest`
function reuses the same `want` variable for both sub-tests, which have
different inputs and expected outputs. The test is configured for
`my-tool` to pass, which consequently causes the `my-array-tool` test to
fail its assertion. The `got` log for the array test confirms that the
tool returns the correct response for the given input.
* **`invoke_my-exec-sql-tool`:** This test fails because the harness
sends the query `SELECT 1`, which is invalid syntax in Firebird.
Subsequent tests for this tool using DDL and DML (`create`, `select`,
`drop`) all pass, confirming its core functionality.
* **`invoke_select-fields-templateParams-tool`:** This test fails due to
a case-sensitivity mismatch in column names (`NAME` vs. `name`). The
expected output in the test harness is hardcoded with a lowercase name.
* **Authentication Tests:** All authentication-related tests are
failing. This appears to be a general issue with the local test
environment setup and is not specific to the Firebird implementation.
## Next Steps
I believe the implementation is ready for review. I am available to make
any requested changes.
Feat #935
---------
Co-authored-by: Yuan Teoh <45984206+Yuan325@users.noreply.github.com>
Co-authored-by: Yuan Teoh <yuanteoh@google.com>
Add options for `RunToolInvokeTest()`, `RunMCPToolCallMethod()`,
`RunExecuteSqlToolInvokeTest()`.
When I was trying to update integration test for Firebird. They have a
different syntax from the usual `SELECT 1` that the other DBs are using.
If we want to update the `SELECT 1` statement in the test cases, we will
have to add a new argument and update all the other integration tests.
With `Options`, we can just add customization without having the need to
update all other integration tests. This allow the integration tests to
be more flexible. With this, we can also remove the
`GetNonSpannerWant()` since those can be set as default values for the
`Options`.
## Add firestore-update-document tool
Adds a new tool for updating existing documents in Firestore
collections.
__What it does:__
- Updates documents at any path in Firestore
- Supports partial updates with field masks for selective field
modification
- Handles all Firestore data types (strings, numbers, booleans,
timestamps, geopoints, arrays, maps, etc.)
- Supports field deletion using updateMask
- Uses Firestore's native JSON format for type safety
- Can update nested fields within maps using dot notation
__Key parameters:__
- `documentPath`: The path of the document to update
- `documentData`: The document content in Firestore JSON format
- `updateMask`: Optional array of field paths for selective updates
- `returnData`: Optional flag to include updated document in response
__Special features:__
- When `updateMask` is provided, only specified fields are updated
- Can access nested fields with dot notation (e.g., 'address.city',
'user.profile.name')
- Without updateMask, performs a merge operation updating all provided
fields
## Description
---
Update mongodb find one to unmarshal projectpayload
## PR Checklist
---
> Thank you for opening a Pull Request! Before submitting your PR, there
are a
> few things you can do to make sure it goes smoothly:
- [ ] Make sure you reviewed
[CONTRIBUTING.md](https://github.com/googleapis/genai-toolbox/blob/main/CONTRIBUTING.md)
- [ ] Make sure to open an issue as a
[bug/issue](https://github.com/googleapis/langchain-google-alloydb-pg-python/issues/new/choose)
before writing your code! That way we can discuss the change, evaluate
designs, and agree on the general idea
- [ ] Ensure the tests and linter pass
- [ ] Code coverage does not decrease (if any source code was changed)
- [ ] Appropriate docs were updated (if necessary)
- [ ] Make sure to add `!` if this involve a breaking change
🛠️Fixes#1153
## Description
---
Adds support for fetching description for looker dimensions and
measures.
Descriptions give important context to any agent about how to filter and
aggregate.
Adds the `description` DimensionFields and MeasureFields
Handles cases when description is not available.
Adds tests.
## PR Checklist
---
- [X] Make sure you reviewed
[CONTRIBUTING.md](https://github.com/googleapis/genai-toolbox/blob/main/CONTRIBUTING.md)
- [X] Make sure to open an issue as a
[bug/issue](https://github.com/googleapis/langchain-google-alloydb-pg-python/issues/new/choose)
before writing your code! That way we can discuss the change, evaluate
designs, and agree on the general idea
- [X] Ensure the tests and linter pass
- [X] Code coverage does not decrease (if any source code was changed)
- [X] Appropriate docs were updated (if necessary)
- [X] Make sure to add `!` if this involve a breaking change
🛠️Fixes#1197
---------
Co-authored-by: Magnus Benediktsson <magnus.benediktsson@nordnet.se>
## Add firestore-add-documents tool
Adds a new tool for creating documents in Firestore collections.
__What it does:__
- Adds documents to any Firestore collection
- Auto-generates unique document IDs
- Supports all Firestore data types (strings, numbers, booleans,
timestamps, geopoints, arrays, maps, etc.)
- Uses Firestore's native JSON format for type safety
__Key parameters:__
- `collectionPath`: Where to add the document
- `documentData`: The document content in Firestore JSON format
- `returnData`: Optional flag to include created document in response
---------
Co-authored-by: Averi Kitsch <akitsch@google.com>
Add optional `dry_run` parameter to bigquery-execute-sql, which defaults
to `false`.
When the `dry_run` parameter is set to `true`, the tool returns the
metadata from the dry run instead of executing the query.
Fixes#703
---------
Co-authored-by: Yuan Teoh <45984206+Yuan325@users.noreply.github.com>
This PR supports TiDB in:
1. sources - tidb: As a data source;
2. tools - tidb - tidbsql: As a prepared SQL running tool;
3. tools - tidb - tidbexecutesql: As an arbitrary SQL running tool (for
development purposes).
And its corresponding docs.
---------
Co-authored-by: Yuan Teoh <45984206+Yuan325@users.noreply.github.com>
Co-authored-by: Yuan Teoh <yuanteoh@google.com>
This is to make bigquery-sql consistent with bigquery-execute-sql. May
not be necessary to have.
- Added a dry run step to identify the query type (e.g., SELECT, DML),
which allows the tool to correctly handle the query's output.
- The recommended high-level client, cloud.google.com/go/bigquery, does
not expose the statement type from a dry run. To circumvent this
limitation, the low-level BigQuery REST API client
(google.golang.org/api/bigquery/v2) was added to gain access to these
necessary details.
---------
Co-authored-by: Averi Kitsch <akitsch@google.com>
Fixes#861
This PR adds support for DuckDB which is a free, open-source, embedded,
in-process, relational database management system (RDBMS) designed for
analytical processing (OLAP)
---------
Co-authored-by: Averi Kitsch <akitsch@google.com>
This pull request introduces a new tool, `neo4j-schema`, for extracting
and processing comprehensive schema information from Neo4j databases. It
includes updates to the documentation, implementation of caching
mechanisms, helper utilities for schema transformation, and
corresponding unit tests. The most important changes are grouped by
theme below:
### Tool Integration:
- **`cmd/root.go`**: Added import for the new `neo4j-schema` tool to
integrate it into the application.
### Documentation:
- **`docs/en/resources/tools/neo4j/neo4j-schema.md`**: Added detailed
documentation for the `neo4j-schema` tool, describing its functionality,
caching behavior, and usage examples.
### Caching Implementation:
- **`internal/tools/neo4j/neo4jschema/cache/cache.go`**: Implemented a
thread-safe, in-memory cache with expiration and optional janitor for
cleaning expired items.
### Unit Tests:
- **`internal/tools/neo4j/neo4jschema/cache/cache_test.go`**: Added
comprehensive tests for the caching system, including functionality for
setting, retrieving, expiration, janitor cleanup, and concurrent access.
### Helper Utilities:
- **`internal/tools/neo4j/neo4jschema/helpers/helpers.go`**: Added
utility functions for processing schema data, including support for APOC
and native Cypher queries, and converting raw query results into
structured formats.
---------
Co-authored-by: Yuan Teoh <45984206+Yuan325@users.noreply.github.com>
- Users have the preference to choose their clients. Below example is
using Gemini CLI.
- Users can use the pre-built Dataplex tools by creating a settings.json
file under .gemini directory. The contents of settings.json would be as
follows:-
```
{
"mcpServers": {
"dataplex": {
"command": "./toolbox",
"args": ["--prebuilt","dataplex","--stdio"],
"env": {
"DATAPLEX_PROJECT": "test-project"
}
}
}
}
```
Fixes#831
---------
Co-authored-by: Wenxin Du <117315983+duwenxin99@users.noreply.github.com>
Co-authored-by: Mateusz Nowak <matnow@google.com>
Co-authored-by: Mateusz Nowak <kontakt@mateusznowak.pl>
Example:
```
alloydb-operations-get:
kind: wait-for-operation
source: alloydb-api-source
method: GET
path: /v1/projects/{{.projectId}}/locations/{{.locationId}}/operations/{{.operationId}}
description: "Makes API call to check whether operation is done or not. This tool is run first then wait for tool. if its still in create phase trigger it after 3 minutes. Print a message saying still not done. We will notify once its done."
pathParams:
- name: projectId
type: string
description: The dynamic path parameter
- name: locationId
type: string
description: The dynamic path parameter
default: us-central1
- name: operationId
type: string
description: Operation status check for previous task
```
This pull request introduces a new tool for executing arbitrary Cypher
queries against a Neo4j database (`neo4j-execute-cypher`) and implements
robust query classification functionality to distinguish between read
and write operations. The changes include updates to documentation, the
addition of a query classifier, and comprehensive test coverage for the
classifier.
### Addition of `neo4j-execute-cypher` tool:
- **Documentation**: Added a new markdown file `neo4j-execute-cypher.md`
that explains the tool's functionality, usage, and configuration
options, including the ability to enforce read-only mode for security.
- **Import statement**: Registered the new tool in the `cmd/root.go`
file to make it available in the toolbox.
### Query classification functionality:
- **Query classifier implementation**: Added `QueryClassifier` in
`classifier.go`, which classifies Cypher queries into read or write
operations based on keywords, procedures, and subquery analysis. It
supports handling edge cases like nested subqueries, multi-word
keywords, and invalid syntax.
- **Test coverage**: Created extensive tests in `classifier_test.go` to
validate the classifier's behavior across various query types, including
abuse cases, subqueries, and procedure calls. Tests ensure the
classifier is robust and does not panic on malformed queries.
---------
Co-authored-by: Yuan Teoh <45984206+Yuan325@users.noreply.github.com>
**Package Renaming**: The `neo4j` package has been renamed to
`neo4jcypher` to better reflect its functionality. This change affects
the tool's implementation
(`internal/tools/neo4j/neo4jcypher/neo4jcypher.go`) and its associated
tests (`internal/tools/neo4j/neo4jcypher/neo4jcypher_test.go`).
- Added a dry run step to identify the query type (e.g., SELECT, DML),
which allows the tool to correctly handle the query's output.
- The recommended high-level client, cloud.google.com/go/bigquery, does
not expose the statement type from a dry run. To circumvent this
limitation, the low-level BigQuery REST API client
(google.golang.org/api/bigquery/v2) was added to gain access to these
necessary details.
fixes: #915
---------
Co-authored-by: Yuan Teoh <45984206+Yuan325@users.noreply.github.com>
Firestore is a NoSQL document database built for automatic scaling, high
performance, and ease of application development. It's a fully managed,
serverless database that supports mobile, web, and server development.
This change adds Firestore as a source in toolbox
Update `tool.Invoke()` to return type `any` instead of `[]any`.
Toolbox return a map with the `results` key, and the SDK reads the
string from the key. So this won't break existing SDK implementation.
Fixes#870
`
tools:
wait_for_tool:
kind: wait-for
description: "Tool puts chat to sleep for given timeout example 3m,10s,
5m etc. For tasks such as cluster creation no need to wait. It takes
longer time such as 8m for instance creation. Use timeout value as
default if nothing is provided"
timeout: 10s
`
Output:
`
prernakakkar@prernakakkar:~/senseai/genai-toolbox$ curl -X POST -H
"Content-Type: application/json" -d '{"duration": "5m"}'
http://127.0.0.1:5000/api/tool/wa
it_for_tool/invoke
{"result":"[\"Wait for 5m0s completed successfully.\"]"}
`
- Spanner: convert array to typed before querying, as Spanner does not
accept non-typed array []any
- BigQuery: fix https://github.com/googleapis/genai-toolbox/issues/793
- Bigtable: add the required `ElemType` for array-type params
- Redis/Valkey: change indexing to append to avoid extra spaces
Add integration tests for array parameters, skipped for the sources not
supporting arrays:
- SQLite
- Cloud SQL MSSQL
- Cloud SQL MySQL
- MSSQL
- MySQL
Allow Toolbox server to automatically update when users modify their
tool configuration file(s), instead of requiring a restart.
This feature is automatically enabled, but can be turned off with the
flag `--disable-reload`.
Optional projectID parameter enables dynamic, cross-project resource
access in BigQuery tools.
This allows a single tool configuration to target different projects at
runtime, rather than being fixed to the project in its source
configuration.
---------
Co-authored-by: Yuan Teoh <45984206+Yuan325@users.noreply.github.com>
Co-authored-by: Wenxin Du <117315983+duwenxin99@users.noreply.github.com>
This feature includes the following:
* Implement initialize lifecycle (including version negotiation)
* Add the v20250326 schema
* Supporting the `DELETE` and `GET` endpoint for MCP.
* Supporting streamable HTTP (without SSE).
* Terminating sessions after timeout (default = 10 minutes from last
active).
* Toolbox do not support batch request. Will response with `Invalid
requests` if batch requests is received.
Add a `default` field to parameters, that enables users to specify a
default value.
e.g.
```
parameters:
- name: name
type: string
default: "some-default-value"
description: The name of the hotel.
```
if this parameter is invoked without specifying `name`, the parameter
would default to "some-default-value"
For parameter manifest, there will be an additional `Required` field.
The default `Required` field is true. If a `default` value is presented,
`Required: false`. Array parameter's item's `Required` field will
inherit the array's `Required` field.
Fixes#475
