MFS Multi-source File-like Search

https://github.com/zilliztech/mfs
Admin
MFS is a context harness for AI agents and for building them, providing one unified workspace over...
Tokens:196,068
Snippets:2,473
Trust Score:8.4
License:Apache-2.0
Update:3 months ago
Show doc for...
Tokens:
Raw
### Create Quickstart Directory and Files

Source: https://github.com/zilliztech/mfs/blob/main/docs/getting-started.md

Sets up a directory structure and creates two markdown files for the MFS quickstart example. This prepares the data to be indexed.

```bash
mkdir -p /tmp/mfs-quickstart/notes
cat > /tmp/mfs-quickstart/README.md <<'EOF'
# MFS quickstart notes

MFS uses a Rust CLI and a Python server.
The default server listens on 127.0.0.1:13619.
EOF

cat > /tmp/mfs-quickstart/notes/search.md <<'EOF'
# Search checklist

Use mfs search when words may not match exactly.
Use mfs grep when the literal token matters.
Use mfs cat with --range to verify exact context.
EOF
```

--------------------------------

### Run MFS Server Setup and Run

Source: https://github.com/zilliztech/mfs/blob/main/docs/getting-started.md

Installs the mfs-server using uv and then sets up the server configuration. Finally, it runs the server. Ensure you have uv installed.

```bash
uv tool install mfs-server
mfs-server setup
mfs-server run
```

--------------------------------

### Manual Sequence for Open Tag Skill Setup

Source: https://github.com/zilliztech/mfs/blob/main/examples/open-tag-skill/README.md

This sequence outlines the manual steps to install and run the MFS server, index data into memory, configure environment variables, and start the Slack agent. Use this in an isolated Slack channel on a non-production machine.

```bash
# 1. install + run MFS
uv tool install mfs-server && mfs-server run            # binds 127.0.0.1:13619

# 2. index your Memory (or just let the mfs-ingest skill do it)
mfs add /path/to/your/repo
mfs add slack://team-memory --config ./slack.toml       # optional Slack history

# 3. point the bot at the permitted scopes + Slack app + backend
export MFS_ALLOWED_SCOPES="file://local/path/to/your/repo,slack://team-memory"
export SLACK_APP_TOKEN="xapp-…"  SLACK_BOT_TOKEN="xoxb-…"
export OPENTAG_BACKEND="claude"  SLACK_CHANNEL_ID="C0…"
export MFS_URL="http://127.0.0.1:13619"  MFS_TOKEN="$(cat ~/.mfs/server.token)"

# 4. preflight, then start the bridge
cd examples/open-tag-skill/open-tag
python scripts/opentag_doctor.py --channel-id "$SLACK_CHANNEL_ID"
uv run --with slack-bolt python scripts/slack_socket_agent.py --backend "$OPENTAG_BACKEND"
```

--------------------------------

### Run Full Server Setup Wizard

Source: https://github.com/zilliztech/mfs/blob/main/docs/configuration.md

Execute the complete server setup wizard to configure all necessary components.

```bash
mfs-server setup
```

--------------------------------

### Run MFS Server Setup and Execution

Source: https://github.com/zilliztech/mfs/blob/main/CONTRIBUTING.md

Run the MFS server setup to create configuration files and then start the server. The server runs on 127.0.0.1:13619 by default.

```bash
uv run mfs-server setup          # optional: write $MFS_HOME/server.toml (Enter through for local defaults)
uv run mfs-server run            # 127.0.0.1:13619
```

--------------------------------

### Install mfs-server with All Connectors

Source: https://github.com/zilliztech/mfs/blob/main/server/python/README.md

Installs the mfs-server along with the SDKs for every available connector.

```bash
uv tool install "mfs-server[all-connectors]"
```

--------------------------------

### Cat API Example

Source: https://github.com/zilliztech/mfs/blob/main/sdks/python/docs/BrowseApi.md

Demonstrates how to use the cat method from the BrowseApi to retrieve file content. Includes setup for authentication and API client configuration.

```python
import mfs_sdk
from mfs_sdk.models.cat_response import CatResponse
from mfs_sdk.rest import ApiException
from pprint import pprint

# Defining the host is optional and defaults to http://localhost
# See configuration.py for a list of all supported configuration parameters.
configuration = mfs_sdk.Configuration(
    host = "http://localhost"
)

# The client must configure the authentication and authorization parameters
# in accordance with the API server security policy.
# Examples for each auth method are provided below, use the example that
# satisfies your auth use case.

# Configure Bearer authorization (opaque): BearerAuth
configuration = mfs_sdk.Configuration(
    access_token = os.environ["BEARER_TOKEN"]
)

# Enter a context with an instance of the API client
with mfs_sdk.ApiClient(configuration) as api_client:
    # Create an instance of the API class
    api_instance = mfs_sdk.BrowseApi(api_client)
    path = 'path_example' # str | 
    range = 'range_example' # str |  (optional)
    meta = False # bool |  (optional) (default to False)
    density = 'density_example' # str |  (optional)
    locator = 'locator_example' # str |  (optional)

    try:
        # Cat
        api_response = api_instance.cat(path, range=range, meta=meta, density=density, locator=locator)
        print("The response of BrowseApi->cat:\n")
        pprint(api_response)
    except Exception as e:
        print("Exception when calling BrowseApi->cat: %s\n" % e)
```

--------------------------------

### Run mfs Server with Specific Bind Address

Source: https://github.com/zilliztech/mfs/blob/main/docs/server.md

Examples of running the mfs server with explicit bind addresses for the API and worker processes. Also shows how to start a detached server process.

```bash
uv run mfs-server run --bind 127.0.0.1:13619
uv run mfs-server api --bind 127.0.0.1:13619
uv run mfs-server worker --concurrency auto
mfs serve start
```

--------------------------------

### Web Connector Configuration Example

Source: https://github.com/zilliztech/mfs/blob/main/docs/connectors/web.md

This TOML configuration specifies the starting point for the crawl, the domains that are allowed to be crawled, and a limit on the number of pages to fetch.

```toml
start_urls = ["https://docs.example.com/"]
allowed_domains = ["docs.example.com"]
max_pages = 100
```

--------------------------------

### Install MFS Server Core

Source: https://github.com/zilliztech/mfs/blob/main/docs/providers.md

Installs the base MFS server with the OpenAI SDK and default ONNX embedding stack. This is the minimal installation.

```bash
uv tool install mfs-server
```

--------------------------------

### Install mfs-server with Specific Connectors

Source: https://github.com/zilliztech/mfs/blob/main/server/python/README.md

Installs the mfs-server with specific connector SDKs, such as PostgreSQL, S3, and Slack.

```bash
uv tool install "mfs-server[pg,s3,slack]"
```

--------------------------------

### Start Local MFS Server

Source: https://github.com/zilliztech/mfs/blob/main/README.md

Use `mfs serve start` to launch a local MFS server process.

```bash
mfs serve start
```

--------------------------------

### Run Specific Setup Wizard Sections

Source: https://github.com/zilliztech/mfs/blob/main/docs/configuration.md

Execute individual sections of the setup wizard, such as 'embedding' or 'auth', for targeted configuration.

```bash
mfs-server setup --section embedding
```

```bash
mfs-server setup --section auth
```

--------------------------------

### Setup LLM Provider Sections via CLI

Source: https://github.com/zilliztech/mfs/blob/main/docs/providers.md

Run these commands to configure the description (for images) and summary (for directories/files) sections using the MFS server setup wizard.

```bash
mfs-server setup --section description   # images
```

```bash
mfs-server setup --section summary       # directories / files
```

--------------------------------

### Configure Embedding Provider via Setup Wizard

Source: https://github.com/zilliztech/mfs/blob/main/docs/providers.md

Launches the MFS setup wizard to configure the embedding section. This is the recommended way to switch providers.

```bash
mfs-server setup --section embedding
```

--------------------------------

### Install LangGraph and MFS SDK

Source: https://github.com/zilliztech/mfs/blob/main/docs/integrations/langgraph.md

Install the necessary libraries for LangGraph and MFS integration. Ensure the MFS Python SDK is also installed separately.

```bash
pip install langgraph langchain-core
# plus the MFS Python SDK (mfs_sdk) — see the SDKs page
```

--------------------------------

### Install MFS Server with Ollama Provider

Source: https://github.com/zilliztech/mfs/blob/main/docs/providers.md

Installs the MFS server with additional dependencies for the Ollama embedding provider. Requires `uv tool install "mfs-server[ollama]"` or `all-providers`.

```bash
uv tool install "mfs-server[ollama]"
```

--------------------------------

### Install LangChain Core and MFS SDK

Source: https://github.com/zilliztech/mfs/blob/main/docs/integrations/langchain.md

Install the necessary Python packages for LangChain core and the MFS SDK. Ensure the MFS SDK is installed from its source location.

```bash
pip install langchain-core
# then install the checked-in MFS SDK from sdks/python (see the SDKs page)
```

--------------------------------

### Install and Run MFS Server with Local Defaults

Source: https://github.com/zilliztech/mfs/blob/main/docs/configuration.md

Install the MFS server, set up its default configuration, and then run it. Use 'mfs status' and 'mfs config show' in another terminal to verify.

```bash
uv tool install mfs-server
mfs-server setup
mfs-server run
```

```bash
mfs status
mfs config show
```

--------------------------------

### Postgres DSN Example

Source: https://github.com/zilliztech/mfs/blob/main/docs/connectors/postgres.md

Provides an example of a standard PostgreSQL connection string (DSN).

```text
postgresql://user:pass@host:5432/dbname
```

--------------------------------

### Start MFS Server from Source

Source: https://github.com/zilliztech/mfs/blob/main/skills/mfs-ingest/SKILL.md

If the `mfs status` command returns 'connection refused', the server may be down. Start it from source using these commands.

```bash
git clone https://github.com/zilliztech/mfs.git && cd mfs/server/python && uv sync && uv run mfs-server setup && uv run mfs-server run
```

--------------------------------

### Install MFS Server with Gemini Provider

Source: https://github.com/zilliztech/mfs/blob/main/docs/providers.md

Installs the MFS server with additional dependencies for the Gemini embedding provider. Requires `uv tool install "mfs-server[gemini]"` or `all-providers`.

```bash
uv tool install "mfs-server[gemini]"
```

--------------------------------

### Example Postgres TOML Configuration

Source: https://github.com/zilliztech/mfs/blob/main/skills/mfs-ingest/reference/connectors/postgres.md

An example TOML file demonstrating DSN, schemas, cursor column, max read rows, and object configurations.

```toml
# in /tmp/mfs-pg-prod.toml
dsN = "env:PG_PROD_DSN"
schemas = ["public", "reporting"]
cursor_column = "updated_at"
max_read_rows = 500000

[[objects]]
match = "/public/tickets"
text_fields = ["title", "description"]
locator_fields = ["id"]
```

--------------------------------

### Install and Run MFS Server Locally

Source: https://github.com/zilliztech/mfs/blob/main/docs/deployment.md

Installs the base MFS server or with all connectors. Optionally sets up the configuration file and runs the server on 127.0.0.1:13619. Defaults to local configurations.

```bash
uv tool install mfs-server                       # base server
uv tool install "mfs-server[all-connectors]"     # + every connector's SDK

mfs-server setup                     # optional: write $MFS_HOME/server.toml
mfs-server run                       # binds 127.0.0.1:13619
```

--------------------------------

### Install MFS Server with All Providers

Source: https://github.com/zilliztech/mfs/blob/main/docs/providers.md

Installs the MFS server with dependencies for Gemini, Voyage, Mistral, Ollama, and Anthropic providers. Does not include the 'local' provider due to its larger dependency stack.

```bash
uv tool install "mfs-server[all-providers]"
```

--------------------------------

### Install MFS Server with Mistral Provider

Source: https://github.com/zilliztech/mfs/blob/main/docs/providers.md

Installs the MFS server with additional dependencies for the Mistral embedding provider. Requires `uv tool install "mfs-server[mistral]"` or `all-providers`.

```bash
uv tool install "mfs-server[mistral]"
```

--------------------------------

### Install Connector Extras

Source: https://github.com/zilliztech/mfs/blob/main/skills/mfs-ingest/reference/troubleshooting.md

If a connector fails due to missing dependencies, use this command to install the necessary extras for your connector.

```bash
uv sync --extra slack
```

--------------------------------

### Install and Build TypeScript SDK

Source: https://github.com/zilliztech/mfs/blob/main/sdks/typescript/README.md

Run these commands in the project directory after cloning to install dependencies and build the TypeScript source code. Node.js and npm are required.

```bash
npm install
npm run build
```

--------------------------------

### Browse API Cat Example

Source: https://github.com/zilliztech/mfs/blob/main/sdks/python/README.md

Example demonstrating how to use the BrowseApi's cat function to read a file. Configure the MFS server URL and optionally set an authorization token. Handles potential API exceptions.

```python
import os

import mfs_sdk
from mfs_sdk.rest import ApiException
from pprint import pprint

# For normal mfs-server run/api, use http://127.0.0.1:13619.
# The generated default http://127.0.0.1:8765 is for generated/smoke runs.
# See configuration.py for a list of all supported configuration parameters.
configuration = mfs_sdk.Configuration(
    host = os.getenv("MFS_URL", "http://127.0.0.1:13619")
)



# Enter a context with an instance of the API client
with mfs_sdk.ApiClient(configuration) as api_client:
    token = os.getenv("MFS_TOKEN")
    if token:
        api_client.set_default_header("Authorization", f"Bearer {token}")

    # Create an instance of the API class
    api_instance = mfs_sdk.BrowseApi(api_client)
    path = os.getenv("MFS_TARGET", "/tmp/mfs_sdk_fixture/auth.md") # str | 
    read_range = None # str | for example "1:120" (optional) 
    meta = False # bool |  (optional) (default to False)
    density = None # str | for example "normal" (optional)

    try:
        # Cat
        api_response = api_instance.cat(path, range=read_range, meta=meta, density=density)
        print("The response of BrowseApi->cat:\n")
        pprint(api_response)
    except ApiException as e:
        print("Exception when calling BrowseApi->cat: %s\n" % e)

```

--------------------------------

### Zendesk TOML Configuration Example

Source: https://github.com/zilliztech/mfs/blob/main/skills/mfs-ingest/reference/connectors/zendesk.md

Example TOML configuration for the Zendesk connector, specifying subdomain, email, API token, and optional read limits.

```toml
subdomain = "acme"
email = "alice@acme.com"
api_token = "env:ZENDESK_API_TOKEN"
max_read_rows = 50000
```

--------------------------------

### MFS Grep Command Examples

Source: https://github.com/zilliztech/mfs/blob/main/evaluation/code-search/experiment-skills/A1_mfs_search.md

Illustrates the usage of `mfs grep` for server-side keyword or full-text search within a specified path. Includes an example for JSON output.

```bash
mfs grep "ERR_TOKEN_EXPIRED" ./src/

```

```bash
mfs --json grep "Middleware" ./src/

```

```bash
mfs grep "todo" ./src/

```

--------------------------------

### TOML Connector Configuration Example

Source: https://github.com/zilliztech/mfs/blob/main/skills/mfs-ingest/SKILL.md

Example of a TOML file used for MFS server connector configuration. Values can be direct strings or environment variable references.

```toml
# mfs-server connector config — <scheme>
# URI: <uri>
<field1> = "<value or env:VAR>"
<field2> = "<value>"
...
```

--------------------------------

### MFS Search Results Example

Source: https://github.com/zilliztech/mfs/blob/main/docs/examples.md

Example output from an `mfs search --all` command, showing results from Slack, Jira, and local files with associated scores and context.

```text
slack://acme/channels/oncall__C0A1B2/messages.jsonl  score=0.91
  [Mon 22:14] @alice: ratelimiter pegged 500ms p99 tail, dump attached
jira://acme/projects/PLAT/issues.jsonl  score=0.83
  PLAT-491  "rate-limit guard misfires under burst"  state=In Progress
file://local/repo/src/throttle.go  score=0.71
  42  func handleRateLimit(req Request) error {
```

--------------------------------

### GitHub Connector Configuration Example

Source: https://github.com/zilliztech/mfs/blob/main/skills/mfs-ingest/reference/connectors/github.md

Example TOML configuration for the GitHub connector. Set the token using an environment variable and specify the branch and metadata indexing options.

```toml
token = "env:GITHUB_TOKEN"
branch = "main"
index_meta = true   # also index issues + PRs
max_read_rows = 10000
```

--------------------------------

### Basic mfs CLI Quickstart Commands

Source: https://github.com/zilliztech/mfs/blob/main/cli/README.md

Demonstrates essential mfs CLI commands for checking server status, adding a local directory for indexing, and performing a hybrid semantic and literal search. Assumes the server is running on the default endpoint.

```bash
mfs status                              # server up? connectors registered?
mfs add ./my-repo                       # queue a local directory for indexing
mfs job show JOB_ID                     # wait until status is succeeded

mfs search "rate limit handler" ./my-repo --top-k 5
mfs cat ./my-repo/src/throttle.go --range 42:78
```

--------------------------------

### JobResponse Model Example

Source: https://github.com/zilliztech/mfs/blob/main/sdks/typescript/docs/JobResponse.md

Demonstrates how to create, log, stringify, and parse a JobResponse object using TypeScript. Ensure you have the '@mfs/sdk' types installed.

```typescript
import type { JobResponse } from '@mfs/sdk'

// TODO: Update the object below with actual values
const example = {
  "id": null,
  "status": null,
  "opKind": null,
  "trigger": null,
  "error": null,
  "totalObjects": null,
  "succeededObjects": null,
  "failedObjects": null,
  "cancelledObjects": null,
  "startedAt": null,
  "finishedAt": null,
} satisfies JobResponse

console.log(example)

// Convert the instance to a JSON string
const exampleJSON: string = JSON.stringify(example)
console.log(exampleJSON)

// Parse the JSON string back to an object
const exampleParsed = JSON.parse(exampleJSON) as JobResponse
console.log(exampleParsed)
```

--------------------------------

### Credential Setup Walkthrough

Source: https://github.com/zilliztech/mfs/blob/main/examples/open-tag-skill/README.md

Request a walkthrough for setting up Slack credentials, including creating a Slack app, enabling Socket Mode, and obtaining bot and app tokens with necessary scopes.

```plain-language
I want to run an Open Tag bot but I don't have any Slack credentials yet. Walk
me through creating the Slack app, turning on Socket Mode, and getting the bot
and app tokens — tell me which scopes to add and where each token goes.
```

--------------------------------

### Python SDK Setup and Server Info

Source: https://github.com/zilliztech/mfs/blob/main/sdks/python/docs/README.md

Demonstrates how to set up the MFS Python SDK client, including configuration with a host URL and bearer token authentication. It then shows how to retrieve and print the server version.

```python
import os

import mfs_sdk


configuration = mfs_sdk.Configuration(
    host=os.getenv("MFS_URL", "http://127.0.0.1:13619")
)

with mfs_sdk.ApiClient(configuration) as api_client:
    token = os.getenv("MFS_TOKEN")
    if token:
        api_client.set_default_header("Authorization", f"Bearer {token}")

    server = mfs_sdk.ServerApi(api_client)
    print(server.get_server_info().version)
```

--------------------------------

### MFS Cat Command Examples

Source: https://github.com/zilliztech/mfs/blob/main/evaluation/code-search/experiment-skills/A3_mfs_search_and_browse.md

Retrieve file content using `mfs cat`. Options include fetching a specific range of lines with `--range`, obtaining structured data with `--locator`, viewing only the heading/signature with `--peek`, or getting headings and the first paragraph of each section with `--skim`.

```bash
mfs cat ./docs/auth.md --peek
```

```bash
mfs cat ./docs/auth.md --skim
```

```bash
mfs cat ./src/handler.py --range 40:80
```

```bash
mfs cat ./src/handler.py --locator '{"lines":[40,80]}'
```

--------------------------------

### Verify MFS CLI installation

Source: https://github.com/zilliztech/mfs/blob/main/docs/getting-started.md

Checks the installed version of the MFS CLI. Run this command after installation to confirm it is accessible.

```bash
mfs --version
```

--------------------------------

### Example Connector Usage

Source: https://github.com/zilliztech/mfs/blob/main/docs/connectors.md

Demonstrates how to use different connectors with common MFS commands like search, ls, and cat. Each command targets a specific data source using its unique URI scheme.

```bash
mfs search "release checklist" github://owner/repo
mfs ls slack://acme/channels
mfs cat postgres://prod-db/public/tickets/rows.jsonl --locator '{"id": 12345}'
```

--------------------------------

### Run MFS Server Setup Wizard

Source: https://github.com/zilliztech/mfs/blob/main/README.md

Initiates the interactive walkthrough for configuring the MFS server. Use `--section` to re-run specific configuration steps.

```bash
mfs-server setup                      # full interactive walkthrough
mfs-server setup --section embedding  # re-run just one section
```

--------------------------------

### Install MFS CLI using Cargo

Source: https://github.com/zilliztech/mfs/blob/main/docs/getting-started.md

Installs the MFS CLI from crates.io using the Cargo package manager. This is an alternative installation method.

```bash
cargo install mfs-cli
```

--------------------------------

### Install mfs-find Skill

Source: https://github.com/zilliztech/mfs/blob/main/examples/deep-research-skill/README.md

Install the mfs-find skill, which is a dependency for the Deep Research skill. This command also specifies AI models and global installation.

```bash
npx skills add zilliztech/mfs --skill mfs-find -a claude-code -a codex -g
```

--------------------------------

### TypeScript SDK Setup and Server Info

Source: https://github.com/zilliztech/mfs/blob/main/sdks/typescript/docs/README.md

Demonstrates how to set up the TypeScript SDK client, configure it with a base path and optional bearer token for authentication, and retrieve server information. Ensure the MFS_URL and MFS_TOKEN environment variables are set for custom configurations.

```typescript
import { Configuration, ServerApi } from "@mfs/sdk";

const basePath = process.env.MFS_URL ?? "http://127.0.0.1:13619";
const token = process.env.MFS_TOKEN;

const configuration = new Configuration({
  basePath,
  headers: token ? { Authorization: `Bearer ${token}` } : undefined,
});

async function main() {
  const server = new ServerApi(configuration);
  const info = await server.getServerInfo();
  console.log(info.version);
}

main().catch((error) => {
  console.error(error);
  process.exitCode = 1;
});
```

--------------------------------

### MFS SDK Basic Operations in Python

Source: https://github.com/zilliztech/mfs/blob/main/docs/sdks.md

Demonstrates initializing the MFS SDK client, setting authentication, and performing server info retrieval, search, file listing, file content retrieval, and ingest job management.

```python
import os

import mfs_sdk


base_url = os.getenv("MFS_URL", "http://127.0.0.1:13619")
token = os.getenv("MFS_TOKEN")
target = os.getenv("MFS_TARGET", "/tmp/mfs_sdk_fixture")

configuration = mfs_sdk.Configuration(host=base_url)

with mfs_sdk.ApiClient(configuration) as client:
    if token:
        client.set_default_header("Authorization", f"Bearer {token}")

    server = mfs_sdk.ServerApi(client)
    retrieval = mfs_sdk.RetrievalApi(client)
    browse = mfs_sdk.BrowseApi(client)
    ingest = mfs_sdk.IngestApi(client)

    print(server.get_server_info().version)

    search = retrieval.search(q="single sign-on login", top_k=3)
    for hit in search.results:
        print(hit.source, hit.score)

    listing = browse.ls(path=target)
    print([entry.name for entry in listing.entries])

    content = browse.cat(path=f"{target}/auth.md")
    print(content)

    job = ingest.add_source(mfs_sdk.AddRequest(target=target, process=False))
    job_status = ingest.get_job(job.job_id)
    print(job_status.id, job_status.status)
```

--------------------------------

### Install mfs CLI with Cargo

Source: https://github.com/zilliztech/mfs/blob/main/cli/README.md

Installs the mfs CLI using Cargo, the Rust package manager. This requires having Rust and Cargo installed on your system.

```bash
cargo install mfs-cli
```

--------------------------------

### Install MFS Server with Local Provider

Source: https://github.com/zilliztech/mfs/blob/main/docs/providers.md

Installs the MFS server with additional dependencies for the local sentence-transformers embedding provider. This installs a larger dependency stack.

```bash
uv tool install "mfs-server[local]"
```

--------------------------------

### Add Slack and Jira, then Search

Source: https://github.com/zilliztech/mfs/blob/main/docs/examples.md

Add Slack and Jira as data sources using their respective configurations. Then, perform a search across all added sources.

```bash
mfs add slack://acme --config ./slack.toml
mfs add jira://acme  --config ./jira.toml
mfs search "why did we revert the burst guard?" --all
```

--------------------------------

### Install mfs CLI with curl

Source: https://github.com/zilliztech/mfs/blob/main/cli/README.md

Installs the mfs CLI using a one-line curl script for Linux and macOS. Ensure you have curl installed and appropriate permissions.

```bash
curl --proto '=https' --tlsv1.2 -LsSf \
  https://github.com/zilliztech/mfs/releases/latest/download/mfs-cli-installer.sh | sh
```

--------------------------------

### Example Help Center Question

Source: https://github.com/zilliztech/mfs/blob/main/evaluation/document-search/README.md

An example of a user-generated question for help-center retrieval.

```text
What is the pricing for email marketing on Wix?
```

--------------------------------

### Add HubSpot and Zendesk, then Search

Source: https://github.com/zilliztech/mfs/blob/main/docs/examples.md

Integrate HubSpot and Zendesk as data sources with their configurations. Subsequently, search for customer-related information across these sources.

```bash
mfs add hubspot://acme  --config ./hubspot.toml
mfs add zendesk://acme  --config ./zendesk.toml
mfs search "why is Globex unhappy with onboarding?" --all
```

--------------------------------

### Install MFS Server with Anthropic Provider

Source: https://github.com/zilliztech/mfs/blob/main/docs/providers.md

Installs the MFS server with additional dependencies for the Anthropic embedding provider. Requires `uv tool install "mfs-server[anthropic]"` or `all-providers`.

```bash
uv tool install "mfs-server[anthropic]"
```

--------------------------------

### Build Docs Locally

Source: https://github.com/zilliztech/mfs/blob/main/docs/development.md

Build the documentation locally using `mkdocs`. Run this command from the repository root.

```bash
uv run --group docs mkdocs build --strict
```

--------------------------------

### Install MFS Server with Voyage Provider

Source: https://github.com/zilliztech/mfs/blob/main/docs/providers.md

Installs the MFS server with additional dependencies for the Voyage embedding provider. Requires `uv tool install "mfs-server[voyage]"` or `all-providers`.

```bash
uv tool install "mfs-server[voyage]"
```

--------------------------------

### Basic MFS SDK Usage with TypeScript

Source: https://github.com/zilliztech/mfs/blob/main/sdks/typescript/README.md

Demonstrates how to initialize the BrowseApi client and perform a 'cat' operation to retrieve file content. It includes configuration for the base path and optional authorization token.

```typescript
import {
  Configuration,
  BrowseApi,
} from '@mfs/sdk';
import type { CatRequest } from '@mfs/sdk';

async function example() {
  const basePath = process.env.MFS_URL ?? "http://127.0.0.1:13619";
  const token = process.env.MFS_TOKEN;
  const api = new BrowseApi(new Configuration({
    basePath,
    headers: token ? { Authorization: `Bearer ${token}` } : undefined,
  }));

  const body = {
    // string
    path: process.env.MFS_TARGET ?? "/tmp/mfs_sdk_fixture/auth.md",
    // string (optional), for example "1:120"
    range: undefined,
    // boolean (optional)
    meta: false,
    // string (optional), for example "normal"
    density: undefined,
  } satisfies CatRequest;

  try {
    const data = await api.cat(body);
    console.log(data);
  } catch (error) {
    console.error(error);
  }
}

// Run the example
example().catch(console.error);
```

--------------------------------

### Install Server Dependencies

Source: https://github.com/zilliztech/mfs/blob/main/CONTRIBUTING.md

Synchronize Python server dependencies using uv. Use '--extra all-connectors' to include all connector SDKs.

```bash
cd server/python
uv sync                          # default deps (local ONNX, Milvus Lite, SQLite)
uv sync --extra all-connectors   # include every connector SDK
```

--------------------------------

### Minimal MFS Quickstart Smoke Test

Source: https://github.com/zilliztech/mfs/blob/main/docs/deployment.md

This script sets up a minimal MFS environment, adds a local file, and performs basic operations like status check, job monitoring, search, and file retrieval. It's useful for a quick end-to-end test.

```bash
mkdir -p /tmp/mfs-quickstart
cat > /tmp/mfs-quickstart/README.md <<'EOF'
MFS uses a Rust CLI and a Python server.
The default server listens on 127.0.0.1:13619.
EOF

export MFS_API_URL=http://127.0.0.1:13619
mfs status

mfs add /tmp/mfs-quickstart          # or: mfs add --upload ... for a container/remote server
mfs job show JOB_ID

mfs search "default endpoint" /tmp/mfs-quickstart --top-k 5
mfs cat /tmp/mfs-quickstart/README.md --range 1:20
```

--------------------------------

### Install Deep Research Skill

Source: https://github.com/zilliztech/mfs/blob/main/examples/deep-research-skill/README.md

Install the Deep Research skill using npx. The --full-depth flag is required as the skill is not in the top-level directory. The -a flags specify AI models to use, and -g indicates global installation.

```bash
npx skills add zilliztech/mfs --full-depth --skill deep-research -a claude-code -a codex -g
```

--------------------------------

### Clone and Install Pre-commit Hooks

Source: https://github.com/zilliztech/mfs/blob/main/CONTRIBUTING.md

Clone the MFS repository and install pre-commit hooks for code formatting and linting.

```bash
git clone https://github.com/zilliztech/mfs.git
cd mfs

# install pre-commit hooks (runs ruff format + cargo fmt on staged files)
pre-commit install
```

--------------------------------

### Postgres Table Structure Example

Source: https://github.com/zilliztech/mfs/blob/main/docs/connectors/postgres.md

Illustrates how MFS organizes tables within schemas, showing rows and schema files.

```text
postgres://prod-db/
└── public/
    ├── tickets/
    │   ├── rows.jsonl     table_rows    → one searchable chunk per row
    │   └── schema.json    table_schema  → browsable schema; searchable with summary enabled
    └── users/
        ├── rows.jsonl
        └── schema.json
```

--------------------------------

### MongoDB Connection URI Examples

Source: https://github.com/zilliztech/mfs/blob/main/docs/connectors/mongo.md

Examples of valid MongoDB connection URIs, including standard and SRV formats.

```text
mongodb://user:pass@host:27017/?authSource=admin
mongodb+srv://user:pass@cluster.mongodb.net/?retryWrites=true
```

--------------------------------

### Export API Key and Add Linear Source

Source: https://github.com/zilliztech/mfs/blob/main/skills/mfs-ingest/reference/connectors/linear.md

Demonstrates exporting the Linear API key as an environment variable and then adding the Linear connector as a source to MFS.

```bash
export LINEAR_API_KEY=lin_api_...
mfs add linear://acme --config /tmp/mfs-linear.toml
```

--------------------------------

### ManifestFile Example

Source: https://github.com/zilliztech/mfs/blob/main/sdks/typescript/docs/ManifestFile.md

Demonstrates how to create a ManifestFile object, convert it to a JSON string, and parse it back. Use this as a template for creating ManifestFile instances.

```typescript
import type { ManifestFile } from '@mfs/sdk'

// TODO: Update the object below with actual values
const example = {
  "path": null,
  "size": null,
  "mtimeNs": null,
  "inode": null,
} satisfies ManifestFile

console.log(example)

// Convert the instance to a JSON string
const exampleJSON: string = JSON.stringify(example)
console.log(exampleJSON)

// Parse the JSON string back to an object
const exampleParsed = JSON.parse(exampleJSON) as ManifestFile
console.log(exampleParsed)
```

--------------------------------

### Install OpenAPI Generator CLI

Source: https://github.com/zilliztech/mfs/blob/main/docs/sdks.md

Installs the OpenAPI generator CLI globally using npm. This is a prerequisite for regenerating SDKs.

```bash
npm install -g @openapitools/openapi-generator-cli
```

--------------------------------

### Update MFS Skills

Source: https://github.com/zilliztech/mfs/blob/main/docs/getting-started.md

Updates installed skills to their latest versions. For project-specific installs, use the 'npx skills add' command again.

```bash
npx skills update
```

--------------------------------

### Run mfs Server from Source

Source: https://github.com/zilliztech/mfs/blob/main/docs/server.md

Commands to set up and run the mfs server locally from its source code. Ensure you are in the 'server/python' directory.

```bash
cd server/python
uv sync --extra all-connectors
uv run mfs-server setup
uv run mfs-server run
```

--------------------------------

### EstimateResponse Example

Source: https://github.com/zilliztech/mfs/blob/main/sdks/typescript/docs/EstimateResponse.md

Demonstrates how to use the EstimateResponse type in TypeScript, including creating an example object, converting it to JSON, and parsing it back.

```typescript
import type { EstimateResponse } from '@mfs/sdk'

// TODO: Update the object below with actual values
const example = {
  "target": null,
  "type": null,
  "objects": null,
  "sampledObjects": null,
  "estChunks": null,
  "estTokens": null,
} satisfies EstimateResponse

console.log(example)

// Convert the instance to a JSON string
const exampleJSON: string = JSON.stringify(example)
console.log(exampleJSON)

// Parse the JSON string back to an object
const exampleParsed = JSON.parse(exampleJSON) as EstimateResponse
console.log(exampleParsed)
```

--------------------------------

### Search Operation with MFS SDK

Source: https://github.com/zilliztech/mfs/blob/main/sdks/typescript/docs/RetrievalApi.md

Illustrates how to perform a search using the MFS SDK's search function. This example shows how to set query parameters like q, path, mode, topK, collapse, and kind. Authentication and valid input are required.

```typescript
import {
  Configuration,
  RetrievalApi,
} from '@mfs/sdk';
import type { SearchRequest } from '@mfs/sdk';

async function example() {
  console.log("🚀 Testing @mfs/sdk SDK...");
  const config = new Configuration({
    // Configure HTTP bearer authorization: BearerAuth
    accessToken: "YOUR BEARER TOKEN",
  });
  const api = new RetrievalApi(config);

  const body = {
    // string
    q: q_example,
    // string (optional)
    path: path_example,
    // 'hybrid' | 'semantic' | 'keyword' (optional)
    mode: mode_example,
    // number (optional)
    topK: 56,
    // boolean (optional)
    collapse: true,
    // string (optional)
    kind: kind_example,
  } satisfies SearchRequest;

  try {
    const data = await api.search(body);
    console.log(data);
  } catch (error) {
    console.error(error);
  }
}

// Run the test
example().catch(console.error);
```

--------------------------------

### EstimateResponse Usage Example

Source: https://github.com/zilliztech/mfs/blob/main/sdks/python/docs/EstimateResponse.md

Example demonstrating how to create and use an EstimateResponse object in Python, including instantiation from JSON and conversion to a dictionary.

```APIDOC
## Example

```python
from mfs_sdk.models.estimate_response import EstimateResponse

# TODO update the JSON string below
json = "{}"
# create an instance of EstimateResponse from a JSON string
estimate_response_instance = EstimateResponse.from_json(json)
# print the JSON string representation of the object
print(EstimateResponse.to_json())

# convert the object into a dict
estimate_response_dict = estimate_response_instance.to_dict()
# create an instance of EstimateResponse from a dict
estimate_response_from_dict = EstimateResponse.from_dict(estimate_response_dict)
```

```

--------------------------------

### Check for Skill Updates

Source: https://github.com/zilliztech/mfs/blob/main/docs/getting-started.md

Checks for available updates for installed skills. Re-run the 'npx skills add' command to update project-level installs.

```bash
npx skills check
```

--------------------------------

### Install MFS Skills Globally

Source: https://github.com/zilliztech/mfs/blob/main/docs/getting-started.md

Installs the mfs-ingest and mfs-find skills for all projects and supported agents. Use without -g for the current project only.

```bash
npx skills add zilliztech/mfs --all -g
```

--------------------------------

### ProbeResponse Usage Example

Source: https://github.com/zilliztech/mfs/blob/main/sdks/python/docs/ProbeResponse.md

Example of how to use the ProbeResponse model in Python, including creating instances from JSON and dictionaries, and converting to JSON and dictionaries.

```python
from mfs_sdk.models.probe_response import ProbeResponse

# Example JSON string (replace with actual JSON data)
json_data = "{}"

# Create an instance of ProbeResponse from a JSON string
probe_response_instance = ProbeResponse.from_json(json_data)

# Print the JSON string representation of the object
print(ProbeResponse.to_json())

# Convert the object into a dict
probe_response_dict = probe_response_instance.to_dict()

# Create an instance of ProbeResponse from a dict
probe_response_from_dict = ProbeResponse.from_dict(probe_response_dict)
```

--------------------------------

### Slack Search Examples

Source: https://github.com/zilliztech/mfs/blob/main/docs/connectors/slack.md

Bash commands demonstrating how to search for specific terms within Slack messages and user data indexed by MFS.

```bash
mfs search "deploy failed" slack://acme/channels/eng-backend__C012345/messages.jsonl
```

```bash
mfs search "Alice Wang" slack://acme/users.jsonl
```

--------------------------------

### Install MFS Skills to Specific Agents

Source: https://github.com/zilliztech/mfs/blob/main/docs/getting-started.md

Installs the mfs-ingest and mfs-find skills to specified agents like claude-code and codex. Supports over 70 agents.

```bash
npx skills add zilliztech/mfs -a claude-code -a codex -g
```

--------------------------------

### MFS Cat Command Examples

Source: https://github.com/zilliztech/mfs/blob/main/skills/mfs-find/SKILL.md

These commands are used to browse and verify content after locating it. '--peek' shows an outline, '--skim' provides outlines with summaries, 'head' shows initial records, and 'tree' displays a subtree structure.

```bash
mfs cat --peek <file>
```

```bash
mfs cat --skim <file>
```

```bash
mfs head -n 20 <uri>
```

```bash
mfs tree <uri> -L 2
```

--------------------------------

### Discord Connector TOML Configuration Example

Source: https://github.com/zilliztech/mfs/blob/main/skills/mfs-ingest/reference/connectors/discord.md

Example TOML configuration for the Discord connector, including token, guild ID, and max read rows.

```toml
token = "env:DISCORD_BOT_TOKEN"
guild_id = "987654321098765432"
max_read_rows = 50000
```
Tokens:
Raw