CI + update docs

Author: solidiquis

.github/workflows/rust_ci.yaml

@@ -60,6 +60,11 @@ jobs:
         with:
           components: clippy, rustfmt
 
+      - name: Install mdbook
+        uses: peaceiris/actions-mdbook@v2
+        with:
+          mdbook-version: "0.4.40"
+
       - name: Run cargo check
         run: cargo check --all-features
 

rust/crates/sift_cli/assets/docs/src/SUMMARY.md

@@ -22,6 +22,7 @@
 # Agentic Tooling
 
 - [MCP Server](./agents/mcp.md)
+  - [Built-in Prompts](./agents/prompts.md)
 - [Agent Skills](./agents/skills.md)
 
 # Reference

rust/crates/sift_cli/assets/docs/src/agents/mcp.md

@@ -34,6 +34,12 @@ not run interactively.
 A typical agent flow is `list_assets` → `list_channels` → `get_data` → `sql`,
 and `upload_dataset` to write results back.
 
+## Built-in prompts
+
+The server also ships [built-in prompts](./prompts.md): ready-made workflows
+that chain these tools to explore an asset, analyze a run, or derive and upload
+a new dataset.
+
 ## Configuring a client
 
 Most MCP clients take a command and arguments. Point yours at the CLI:

rust/crates/sift_cli/assets/docs/src/agents/prompts.md

@@ -0,0 +1,117 @@
+# Built-in Prompts
+
+The Sift MCP server ships with a set of built-in prompts: ready-made workflows
+that walk an agent through a complete task using the [MCP tools](./mcp.md). A
+prompt expands into step-by-step instructions tailored to the arguments you
+give it, so you start a focused session without writing the workflow yourself.
+
+Prompts are served by `sift-cli mcp`, so the server must be
+[configured as a client](./mcp.md#configuring-a-client) first.
+
+## Using a prompt
+
+How you invoke a prompt depends on your MCP client, and on the name you
+registered the server under. Assuming you registered it as `sift` in your
+`.mcp.json`:
+
+```json
+{
+  "mcpServers": {
+    "sift": {
+      "command": "sift-cli",
+      "args": ["mcp"]
+    }
+  }
+}
+```
+
+then in Claude Code each prompt is available as a slash command of the form
+`/mcp__sift__<prompt>`:
+
+- `/mcp__sift__explore_asset`
+- `/mcp__sift__analyze_run`
+- `/mcp__sift__derive_and_upload`
+
+The `sift` in the command is your registered server name, so a different name
+in `.mcp.json` changes the prefix accordingly. Arguments are passed positionally
+after the command, separated by spaces; wrap any argument containing spaces in
+quotes. Other MCP clients discover the same prompts but may present them
+differently, so check your client's documentation for its invocation syntax.
+
+## `explore_asset`
+
+A read-only starting point for a session. It resolves an asset, lists its
+recent runs, and inventories its channels, without pulling any sample data.
+
+| Argument | Required | Description                            |
+| -------- | -------- | -------------------------------------- |
+| `asset`  | yes      | Asset name to explore (exact or partial match). |
+
+Example:
+
+```
+/mcp__sift__explore_asset "Falcon 9 Booster"
+```
+
+The agent resolves the asset, then reports its most recent runs and the channel
+inventory grouped by data type, surfacing the exact run and channel names so you
+can reuse them with the other prompts.
+
+## `analyze_run`
+
+Pulls a run's channel data and produces a per-channel statistical summary. You
+can target specific channels and pose a question for the agent to answer.
+
+| Argument   | Required | Description                                                  |
+| ---------- | -------- | ------------------------------------------------------------ |
+| `asset`    | yes      | Asset the run belongs to.                                    |
+| `run`      | yes      | Run to analyze.                                              |
+| `channels` | no       | Channels to pull. Omit to let the agent choose a subset.     |
+| `question` | no       | A question for the agent to answer from the data.            |
+
+Analyze every channel on a run:
+
+```
+/mcp__sift__analyze_run "Falcon 9 Booster" "Static Fire 2024-05-01"
+```
+
+Target specific channels and ask a question:
+
+```
+/mcp__sift__analyze_run "Falcon 9 Booster" "Static Fire 2024-05-01" "chamber_pressure,fuel_temp" "Did chamber pressure stay within nominal range?"
+```
+
+The agent pulls the data with `get_data`, summarizes it with `sql` (row counts,
+min/max/mean, null rate), and reports the Parquet paths so you can continue the
+work.
+
+## `derive_and_upload`
+
+Derives a new dataset from an existing run via SQL and uploads it back to Sift.
+Because the upload is a write, the agent confirms the destination with you before
+running it.
+
+| Argument       | Required | Description                                                       |
+| -------------- | -------- | ----------------------------------------------------------------- |
+| `source_asset` | yes      | Asset to read from.                                               |
+| `source_run`   | yes      | Run to read from.                                                 |
+| `transform`    | yes      | Plain-language description of the transform to apply.             |
+| `target_asset` | no       | Asset to write to. The agent proposes a default if omitted.       |
+| `target_run`   | no       | Run to create. The agent asks whether to create one if omitted.   |
+
+Derive a new dataset, letting the agent propose the destination:
+
+```
+/mcp__sift__derive_and_upload "Falcon 9 Booster" "Static Fire 2024-05-01" "1-second average of chamber_pressure and fuel_temp"
+```
+
+Specify the destination explicitly:
+
+```
+/mcp__sift__derive_and_upload "Falcon 9 Booster" "Static Fire 2024-05-01" "1-second average of chamber_pressure" "Falcon 9 Booster-derived" "Static Fire 2024-05-01 1s-avg"
+```
+
+The agent extracts the source data, applies the transform with `sql` (keeping
+`timestamp_unix_nanos` as the first column, as Sift requires), confirms the
+target asset, run, and any tags with you, then uploads the result with
+`upload_dataset`.

rust/crates/sift_cli/build.rs

@@ -30,9 +30,6 @@ fn require_mdbook() {
         .unwrap_or(false);
 
     if !installed {
-        panic!(
-            "`mdbook` not found on PATH. Install it with `cargo install mdbook --locked` \
-             or run `make bootstrap`."
-        );
+        panic!("`mdbook` not found on PATH. Install it with `cargo install mdbook --locked`.");
     }
 }