Introduce workflow.jobs support and script path overrides; bump version to 0.1.14

- Add parsing and execution support for `.solstice/workflow.kdl` with job-specific configurations, including `runs_on`, `script path`, and `workflow_job_id`.
- Enable job grouping via `group_id` for cohesive workflow processing.
- Update orchestrator to pass workflow-specific parameters to `cloud-init` for finer control over execution.
- Refactor enqueue logic to handle multiple jobs per workflow with fallback to single job when no workflow is defined.
- Enhance dependencies for workflow parsing by integrating `base64`, `regex`, and `uuid`.
- Increment orchestrator version to 0.1.14 for release.

Signed-off-by: Till Wegmueller <toasterson@gmail.com>

crates/common/src/messages.rs

@@ -11,6 +11,9 @@ pub struct JobRequest {
     pub schema_version: String, // e.g., "jobrequest.v1"
     /// Unique request identifier for idempotency and tracing correlation.
     pub request_id: Uuid,
+    /// Optional group identifier to correlate multiple job runs for the same workflow request.
+    #[serde(default)]
+    pub group_id: Option<Uuid>,
     /// Source system of this request (forge or manual trigger).
     pub source: SourceSystem,
     /// Repository clone URL (SSH or HTTPS).
@@ -24,11 +27,17 @@ pub struct JobRequest {
     /// Commit SHA to check out.
     pub commit_sha: String,
     /// Optional path to the workflow file within the repo (KDL).
+    #[serde(default)]
     pub workflow_path: Option<String>,
     /// Optional specific job id from the workflow to run.
+    #[serde(default)]
     pub workflow_job_id: Option<String>,
     /// Optional scheduling hint selecting a base image or host group.
+    #[serde(default)]
     pub runs_on: Option<String>,
+    /// Optional script path within the repository to execute instead of the default .solstice/job.sh
+    #[serde(default)]
+    pub script_path: Option<String>,
     /// Submission timestamp (UTC).
     pub submitted_at: OffsetDateTime,
 }
@@ -54,6 +63,7 @@ impl JobRequest {
         Self {
             schema_version: default_jobrequest_schema(),
             request_id: Uuid::new_v4(),
+            group_id: None,
             source,
             repo_url: repo_url.into(),
             repo_owner: None,
@@ -62,6 +72,7 @@ impl JobRequest {
             workflow_path: None,
             workflow_job_id: None,
             runs_on: None,
+            script_path: None,
             submitted_at: OffsetDateTime::now_utc(),
         }
     }

crates/forge-integration/Cargo.toml

@@ -24,3 +24,7 @@ futures-util = "0.3"
 # S3/Garage upload
 aws-config = { version = "1", default-features = false, features = ["behavior-version-latest", "rt-tokio"] }
 aws-sdk-s3 = { version = "1", default-features = false, features = ["rt-tokio", "rustls"] }
+# Workflow parsing helpers
+base64 = "0.22"
+regex = "1.11"
+uuid = { version = "1", features = ["v4"] }

crates/forge-integration/src/main.rs

@@ -584,6 +584,8 @@ struct PushPayload {
     repository: RepoInfo,
 }
 
+use base64::Engine;
+
 async fn handle_push(state: Arc<AppState>, body: Bytes) -> StatusCode {
     let payload: PushPayload = match serde_json::from_slice(&body) {
         Ok(p) => p,
@@ -602,29 +604,31 @@ async fn handle_push(state: Arc<AppState>, body: Bytes) -> StatusCode {
     let repo_url = pick_repo_url(&payload.repository);
     let sha = payload.after;
 
-    match enqueue_job(&state, repo_url.clone(), sha.clone(), None).await {
-        Ok(jr) => {
+    match enqueue_jobs(&state, repo_url.clone(), sha.clone(), None).await { 
+        Ok(jobs) => {
             if let (Some(base), Some(token), Some(orch)) = (
                 state.forgejo_base.as_ref(),
                 state.forgejo_token.as_ref(),
                 state.orch_http_base.as_ref(),
             ) {
+                if let Some(first) = jobs.first() {
                     let _ = post_commit_status(
                         base,
                         token,
-                    &jr.repo_url,
-                    &jr.commit_sha,
+                        &first.repo_url,
+                        &first.commit_sha,
                         &state.forge_context,
                         "pending",
-                    Some(&format!("{}/jobs/{}/logs", orch.trim_end_matches('/'), jr.request_id)),
-                    Some("Solstice job queued"),
+                        Some(&format!("{}/jobs/{}/logs", orch.trim_end_matches('/'), first.request_id)),
+                        Some("Solstice jobs queued"),
                     )
                     .await;
                 }
+            }
             StatusCode::ACCEPTED
         }
         Err(e) => {
-            error!(error = %e, "failed to publish job");
+            error!(error = %e, "failed to publish jobs");
             StatusCode::INTERNAL_SERVER_ERROR
         }
     }
@@ -685,10 +689,10 @@ async fn handle_pull_request(state: Arc<AppState>, body: Bytes) -> StatusCode {
         .map(|l| l.name)
         .collect();
 
-    match enqueue_job(&state, repo_url, sha, Some(label_names)).await {
+    match enqueue_jobs(&state, repo_url, sha, Some(label_names)).await {
         Ok(_) => StatusCode::ACCEPTED,
         Err(e) => {
-            error!(error = %e, "failed to publish job");
+            error!(error = %e, "failed to publish jobs");
             StatusCode::INTERNAL_SERVER_ERROR
         }
     }
@@ -710,19 +714,134 @@ fn pick_repo_url_pr(repo: &PrRepoInfo) -> String {
         .to_string()
 }
 
-async fn enqueue_job(state: &Arc<AppState>, repo_url: String, commit_sha: String, labels: Option<Vec<String>>) -> Result<common::JobRequest> {
+struct ParsedJob { id: String, runs_on: Option<String>, script: Option<String> }
+
+async fn fetch_workflow_kdl(base: Option<&str>, token: Option<&str>, owner: &str, repo: &str, sha: &str) -> Result<Option<String>> {
+    // Try Forgejo API: GET /repos/{owner}/{repo}/contents/.solstice/workflow.kdl?ref={sha}
+    if let Some(base) = base {
+        let url = format!("{}/repos/{}/{}/contents/.solstice/workflow.kdl?ref={}", base.trim_end_matches('/'), owner, repo, sha);
+        let client = reqwest::Client::new();
+        let mut req = client.get(&url);
+        if let Some(tok) = token { req = req.bearer_auth(tok); }
+        let resp = req.send().await.into_diagnostic()?;
+        if resp.status().is_success() {
+            let v: serde_json::Value = resp.json().await.into_diagnostic()?;
+            if let Some(enc) = v.get("encoding").and_then(|e| e.as_str()) {
+                if enc.eq_ignore_ascii_case("base64") {
+                    if let Some(content) = v.get("content").and_then(|c| c.as_str()) {
+                        let decoded = base64::engine::general_purpose::STANDARD.decode(content.replace('\n', "")).into_diagnostic()?;
+                        let s = String::from_utf8(decoded).into_diagnostic()?;
+                        return Ok(Some(s));
+                    }
+                }
+            }
+        }
+    }
+    Ok(None)
+}
+
+fn parse_workflow_jobs(kdl: &str) -> Vec<ParsedJob> {
+    // Minimal heuristic parser: find lines starting with `job id="..."` and capture runs_on and an optional `script path="..."` line inside the block.
+    // This is not a full KDL parser but should work for our simple workflows.
+    let mut out = Vec::new();
+    let mut lines = kdl.lines().enumerate().peekable();
+    while let Some((i, line)) = lines.next() {
+        let l = line.trim();
+        if l.starts_with("job ") && l.contains("id=") {
+            // capture id and runs_on on the same line if present
+            let id = capture_attr(l, "id");
+            let mut runs_on = capture_attr(l, "runs_on");
+            let mut script: Option<String> = None;
+            // consume block until closing '}' balancing braces
+            let mut depth = if l.ends_with('{') { 1 } else { 0 };
+            while let Some((_j, ln)) = lines.peek().cloned() {
+                let t = ln.trim();
+                if t.ends_with('{') { depth += 1; }
+                if t.starts_with('}') {
+                    if depth == 0 { break; }
+                    depth -= 1;
+                    if depth == 0 { lines.next(); break; }
+                }
+                // within job block: look for step or script lines; allow `script path="..."` or `step name="..." run="..."`
+                if t.starts_with("script ") && t.contains("path=") {
+                    if let Some(p) = capture_attr(t, "path") { script = Some(p); }
+                }
+                // Also allow runs_on within block as override
+                if t.contains("runs_on=") && runs_on.is_none() {
+                    runs_on = capture_attr(t, "runs_on");
+                }
+                lines.next();
+            }
+            if let Some(id_val) = id {
+                out.push(ParsedJob { id: id_val, runs_on, script });
+            }
+        }
+    }
+    out
+}
+
+fn capture_attr(line: &str, key: &str) -> Option<String> {
+    // Accept key="value" or key='value'
+    let pattern1 = format!("{}=\"", key);
+    if let Some(start) = line.find(&pattern1) {
+        let rest = &line[start + pattern1.len()..];
+        if let Some(end) = rest.find('"') { return Some(rest[..end].to_string()); }
+    }
+    let pattern2 = format!("{}='", key);
+    if let Some(start) = line.find(&pattern2) {
+        let rest = &line[start + pattern2.len()..];
+        if let Some(end) = rest.find('\'') { return Some(rest[..end].to_string()); }
+    }
+    None
+}
+
+async fn enqueue_jobs(state: &Arc<AppState>, repo_url: String, commit_sha: String, labels: Option<Vec<String>>) -> Result<Vec<common::JobRequest>> {
+    use uuid::Uuid;
     if repo_url.is_empty() {
         miette::bail!("missing repo_url in webhook payload");
     }
-    let mut jr = common::JobRequest::new(common::SourceSystem::Forgejo, repo_url, commit_sha);
+
+    // Base request (will be cloned per job when a workflow defines multiple jobs)
+    let mut base = common::JobRequest::new(common::SourceSystem::Forgejo, repo_url, commit_sha);
     // Try to populate repo_owner/repo_name from URL for accurate status routing
-    if let Some((owner, name)) = parse_owner_repo(&jr.repo_url) {
-        jr.repo_owner = Some(owner);
-        jr.repo_name = Some(name);
+    if let Some((owner, name)) = parse_owner_repo(&base.repo_url) {
+        base.repo_owner = Some(owner);
+        base.repo_name = Some(name);
     }
-    // Infer runs_on from repo map, labels, or default
-    jr.runs_on = infer_runs_on(state, &jr.repo_url, labels.as_ref().map(|v| v.as_slice()));
+
+    // Attempt to fetch and parse .solstice/workflow.kdl at the commit
+    let mut published: Vec<common::JobRequest> = Vec::new();
+    if let (Some(owner), Some(repo)) = (base.repo_owner.clone(), base.repo_name.clone()) {
+        if let Ok(Some(kdl)) = fetch_workflow_kdl(state.forgejo_base.as_deref(), state.forgejo_token.as_deref(), &owner, &repo, &base.commit_sha).await {
+            let jobs = parse_workflow_jobs(&kdl);
+            if !jobs.is_empty() {
+                let gid = Uuid::new_v4();
+                for pj in jobs {
+                    let mut jr = base.clone();
+                    jr.request_id = Uuid::new_v4(); // unique per job
+                    jr.group_id = Some(gid);
+                    jr.workflow_path = Some(".solstice/workflow.kdl".to_string());
+                    jr.workflow_job_id = Some(pj.id);
+                    // runs_on precedence: job-specific -> inferred (labels/map/default)
+                    jr.runs_on = pj
+                        .runs_on
+                        .clone()
+                        .or_else(|| infer_runs_on(state, &jr.repo_url, labels.as_ref().map(|v| v.as_slice())));
+                    jr.script_path = pj.script.clone();
+
                     common::publish_job(&state.mq_cfg, &jr).await?;
-    info!(request_id = %jr.request_id, repo = %jr.repo_url, sha = %jr.commit_sha, runs_on = ?jr.runs_on, "enqueued job from webhook");
-    Ok(jr)
+                    info!(request_id = %jr.request_id, group_id = ?jr.group_id, repo = %jr.repo_url, sha = %jr.commit_sha, job = ?jr.workflow_job_id, runs_on = ?jr.runs_on, "enqueued workflow job");
+                    published.push(jr);
+                }
+                return Ok(published);
+            }
+        }
+    }
+
+    // Fallback: no workflow or no jobs parsed — enqueue a single job
+    base.runs_on = infer_runs_on(state, &base.repo_url, labels.as_ref().map(|v| v.as_slice()));
+    common::publish_job(&state.mq_cfg, &base).await?;
+    info!(request_id = %base.request_id, repo = %base.repo_url, sha = %base.commit_sha, runs_on = ?base.runs_on, "enqueued single job (no workflow)");
+    published.push(base);
+    Ok(published)
 }

crates/orchestrator/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "orchestrator"
-version = "0.1.13"
+version = "0.1.14"
 edition = "2024"
 build = "build.rs"
 

crates/orchestrator/src/main.rs

@@ -348,7 +348,15 @@ async fn main() -> Result<()> {
                     disk_gb,
                     network: None, // libvirt network handled in backend
                     nocloud: image.nocloud,
-                    user_data: Some(make_cloud_init_userdata(&job.repo_url, &job.commit_sha, job.request_id, &pubkey_text)),
+                    user_data: Some(make_cloud_init_userdata(
+                                            &job.repo_url,
+                                            &job.commit_sha,
+                                            job.request_id,
+                                            &pubkey_text,
+                                            job.workflow_job_id.as_deref(),
+                                            job.script_path.as_deref(),
+                                            job.group_id,
+                                        )),
                 };
                 if !spec.nocloud {
                     warn!(label = %label_resolved, "image is not marked nocloud=true; cloud-init may not work");
@@ -427,7 +435,14 @@ fn make_cloud_init_userdata(
     commit_sha: &str,
     _request_id: uuid::Uuid,
     ssh_pubkey: &str,
+    workflow_job_id: Option<&str>,
+    script_path: Option<&str>,
+    group_id: Option<uuid::Uuid>,
 ) -> Vec<u8> {
+    let mut extra = String::new();
+    if let Some(j) = workflow_job_id { extra.push_str(&format!("      workflow_job_id: {}\n", j)); }
+    if let Some(s) = script_path { extra.push_str(&format!("      script_path: {}\n", s)); }
+    if let Some(g) = group_id { extra.push_str(&format!("      group_id: {}\n", g)); }
     let s = format!(
         r#"#cloud-config
 users:
@@ -441,10 +456,12 @@ write_files:
     content: |
       repo_url: {repo}
       commit_sha: {sha}
+{extra}
 "#,
         repo = repo_url,
         sha = commit_sha,
         ssh_pubkey = ssh_pubkey.trim(),
+        extra = extra,
     );
     s.into_bytes()
 }
@@ -475,6 +492,9 @@ mod tests {
             "deadbeef",
             req_id,
             "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIEfakepubkey user@example",
+            None,
+            None,
+            None,
         );
         let s = String::from_utf8(data).unwrap();
         assert!(s.contains("#cloud-config"));

crates/workflow-runner/src/main.rs

@@ -45,6 +45,12 @@ struct Opts {
 struct JobFile {
     repo_url: String,
     commit_sha: String,
+    #[serde(default)]
+    workflow_job_id: Option<String>,
+    #[serde(default)]
+    script_path: Option<String>,
+    #[serde(default)]
+    group_id: Option<String>,
 }
 
 async fn read_job_file() -> Result<JobFile> {
@@ -326,8 +332,13 @@ async fn ensure_repo(repo: &str, sha: &str, workdir: &str) -> Result<()> {
     }
 }
 
-async fn run_job_script(workdir: &str) -> Result<i32> {
-    let script = format!("{}/.solstice/job.sh", workdir);
+async fn run_job_script(workdir: &str, script_override: Option<&str>) -> Result<i32> {
+    // Determine the script to execute: prefer override from job.yaml, else default .solstice/job.sh
+    let script = if let Some(path) = script_override {
+        if path.starts_with('/') { path.to_string() } else { format!("{}/{}", workdir, path.trim_start_matches("./")) }
+    } else {
+        format!("{}/.solstice/job.sh", workdir)
+    };
     if !fs::try_exists(&script).await.into_diagnostic()? {
         warn!(path = %script, "job script not found");
         eprintln!("{}", ndjson_line("job_run", "error", &format!("job script not found at {}", script), None));
@@ -458,8 +469,11 @@ async fn main() -> Result<()> {
 
     let code = match ensure_repo(&repo, &sha, &workdir).await {
         Ok(_) => {
+            // Read job.yaml to get optional script override
+            let jf = read_job_file().await.ok();
+            let script_override = jf.as_ref().and_then(|j| j.script_path.as_deref());
             // proceed to run job script
-            run_job_script(&workdir).await?
+            run_job_script(&workdir, script_override).await?
         }
         Err(e) => {
             eprintln!("{}", ndjson_line("env_setup", "error", &format!("failed to prepare repo: {}", e), None));