Add federation worker and canonical handoffs

2026-07-28 16:17:18 +04:00
parent 58793a5aa3
commit 2cecbc4015
22 changed files with 1429 additions and 108 deletions
@@ -75,6 +75,10 @@ type CLIAdapter struct {
 	// release. Nil disables Release (adapters built without one refuse
 	// loudly rather than skip validation).
 	CAS continuity.CAS
+	// Remote, when set by a federation worker, is pushed after the scratch
+	// commit and before the pane claim is released. Git is the cross-machine
+	// transport; a CAS handoff must never point at an unpushed anchor.
+	Remote string
 }

 // HandoffFile is the convention the agent writes its §6.1 handoff to before
@@ -83,8 +87,9 @@ type CLIAdapter struct {
 // uploads the one the agent wrote (herdr does not write handoffs, §6.1).
 const HandoffFile = ".orchestra-handoff.json"

-// HandoffReportFile is the only handoff artifact an opaque harness authors.
-// The worker which owns the checkout derives and seals the canonical JSON.
+// HandoffReportFile holds the agent's small, labelled answer during release.
+// It is not a report: the worker parses it, derives the protocol facts, and
+// seals the resulting canonical JSON.
 const HandoffReportFile = ".orchestra-handoff-report.md"

 func (a CLIAdapter) CreateWorktree(ctx context.Context, repo, root, taskID string) (string, error) {
@@ -149,9 +154,16 @@ func (a CLIAdapter) Bootstrap(ctx context.Context, s Session, ref string) error
 	return a.prompt(ctx, s, fmt.Sprintf(bootstrapPrompt, ref), time.Minute)
 }

-const handoffPrompt = `Orchestra is about to rotate this task to a fresh session (context budget reached).
-Before you stop, write a concise semantic handoff report to ` + HandoffReportFile + ` at the worktree root: what changed, validation evidence, remaining work, review findings, and dead ends/open questions.
-Do not write protocol JSON, git anchors, or file hashes; Orchestra's checkout worker collects and validates those facts. Do not edit TASK.md. Once written, stop normally.`
+const handoffPrompt = `Orchestra is about to rotate this task. Write ONLY the following labelled answers to ` + HandoffReportFile + `, then stop. Output nothing else.
+
+NEXT: the single next action (one line).
+WHY: why that is next (one line).
+REMAINING: outstanding items, one line each. If none: NONE.
+DEAD ENDS: approaches tried that failed — "tried X → failed because Y", one per line. If none: NONE.
+OPEN Q: unresolved decisions, one line each. If none: NONE.
+LEARNED: constraints discovered that are NOT in TASK.md, one line each. If none: NONE.
+
+Do NOT include: what you completed (the diff shows it), the goal or done-criteria (TASK.md holds them), git SHAs/branches/paths, or a prose summary. No headings and no report. Do not edit TASK.md.`

 // RequestHandoff prompts the agent to write HandoffFile before Release reads
 // it. Optional capability: adapters without a live pane (tests, etc.) can
@@ -183,7 +195,7 @@ func (a CLIAdapter) RequestHandoffReason(ctx context.Context, s Session, reason
 	case "milestone":
 		sb.WriteString("A coherent unit of work looks complete (a successful commit). If the next step is independent of what you just did, this is a good point to hand off.\n")
 	}
-	fmt.Fprintf(&sb, "Before you stop, write a concise semantic handoff report to %s at the worktree root (reason: %q)", HandoffReportFile, reason)
+	fmt.Fprintf(&sb, "Before you stop, write the labelled handoff answers requested below to %s at the worktree root (reason: %q).\n\n%s", HandoffReportFile, reason, handoffPrompt[strings.Index(handoffPrompt, "NEXT:"):])
 	if len(deadEnds) > 0 {
 		sb.WriteString(" and a dead_ends entry for each of the following:\n")
 		for _, d := range deadEnds {
@@ -192,7 +204,7 @@ func (a CLIAdapter) RequestHandoffReason(ctx context.Context, s Session, reason
 	} else {
 		sb.WriteString(".\n")
 	}
-	sb.WriteString("Include what changed, validation evidence, remaining work, review findings, and any dead ends. Do not write protocol JSON, git anchors, or file hashes; Orchestra collects those. Do not edit TASK.md. Once written, stop normally.")
+	sb.WriteString("Do not add a prose summary, completed-work narration, or protocol JSON. Once written, stop normally.")
 	return a.prompt(ctx, s, sb.String(), time.Minute)
 }

@@ -234,9 +246,9 @@ func (a CLIAdapter) NotifyConventionsChanged(ctx context.Context, s Session) err
 	return a.prompt(ctx, s, conventionsPrompt, time.Minute)
 }

-// Release reads the §6.1 handoff the agent wrote to HandoffFile at the
-// worktree root, validates its schema and anchor against the worktree's real
-// HEAD, uploads it to CAS, and only then releases herdr's claim on the pane
+// Release reads the semantic report the agent wrote at the worktree root,
+// derives and validates the canonical handoff from the worktree's real Git
+// state, uploads it to CAS, and only then releases herdr's claim on the pane
 // via the real pane.release_agent(pane_id, source, agent) method (confirmed
 // live against herdr, AUDIT.md Phase 0 — the invented "pane.release" never
 // existed and could never have returned a handoff_ref regardless, since
@@ -256,7 +268,7 @@ func (a CLIAdapter) Release(ctx context.Context, s Session) (string, error) {
 	if strings.TrimSpace(string(b)) == "" {
 		return "", fmt.Errorf("adapter: semantic handoff report is empty")
 	}
-	h, err := canonicalHandoff(s, string(b))
+	h, err := canonicalHandoff(s, string(b), a.lastObservedCommand(s))
 	if err != nil {
 		return "", err
 	}
@@ -270,6 +282,20 @@ func (a CLIAdapter) Release(ctx context.Context, s Session) (string, error) {
 			return "", fmt.Errorf("adapter: handoff dirty file changed since it was written: %s", d.Path)
 		}
 	}
+	// The semantic report is transferred in the CAS handoff, not in the
+	// scratch checkout. Keeping it in the scratch commit makes a successor
+	// mistake the predecessor's report for a newly requested handoff and can
+	// cause an immediate release/pickup loop.
+	dirty := h.Anchor.Dirty[:0]
+	for _, d := range h.Anchor.Dirty {
+		if filepath.Clean(d.Path) != HandoffReportFile {
+			dirty = append(dirty, d)
+		}
+	}
+	h.Anchor.Dirty = dirty
+	if err := os.Remove(path); err != nil && !os.IsNotExist(err) {
+		return "", fmt.Errorf("adapter: remove transferred semantic report: %w", err)
+	}
 	// Atomically commit whatever the handoff described as dirty onto a
 	// per-task scratch branch (§6.2 step 3) *before* uploading, so the
 	// successor's pickup validation collapses to a single HEAD compare
@@ -279,6 +305,11 @@ func (a CLIAdapter) Release(ctx context.Context, s Session) (string, error) {
 		if err := continuity.ScratchCommit(s.Worktree, branch, "orchestra: pre-release WIP snapshot ("+h.Meta.ID+")"); err != nil {
 			return "", fmt.Errorf("adapter: scratch commit: %w", err)
 		}
+		if a.Remote != "" {
+			if err := continuity.ScratchPush(s.Worktree, branch, a.Remote); err != nil {
+				return "", fmt.Errorf("adapter: push scratch branch: %w", err)
+			}
+		}
 		newSHA, err := HeadSHA(s.Worktree)
 		if err != nil {
 			return "", fmt.Errorf("adapter: read scratch HEAD: %w", err)
@@ -302,8 +333,13 @@ func (a CLIAdapter) Release(ctx context.Context, s Session) (string, error) {
 }

 // canonicalHandoff keeps Git-derived protocol facts on the worker that owns
-// the checkout. The harness contributes only the semantic report (B17).
-func canonicalHandoff(s Session, report string) (continuity.Handoff, error) {
+// the checkout. Every authored field comes from the validated agent answer;
+// it never fabricates task intent or a circular next action.
+func canonicalHandoff(s Session, answer, command string) (continuity.Handoff, error) {
+	authored, err := parseHandoffAnswer(answer)
+	if err != nil {
+		return continuity.Handoff{}, fmt.Errorf("adapter: invalid handoff answer: %w", err)
+	}
 	sha, err := HeadSHA(s.Worktree)
 	if err != nil {
 		return continuity.Handoff{}, fmt.Errorf("adapter: read worktree HEAD: %w", err)
@@ -317,16 +353,131 @@ func canonicalHandoff(s Session, report string) (continuity.Handoff, error) {
 		return continuity.Handoff{}, err
 	}
 	return continuity.Handoff{
-		Meta:      continuity.Meta{ID: handoffID(s), Reason: "threshold"},
-		Anchor:    continuity.Anchor{GitSHA: sha, Branch: strings.TrimSpace(string(branchOut)), Dirty: dirty},
-		Goal:      "Continue Orchestra task " + s.PaneID,
-		DoneWhen:  []string{"Task completion is reported to Orchestra"},
-		Action:    "Read the semantic handoff report and continue the task.",
-		Command:   "cat " + HandoffReportFile,
-		Remaining: []string{report},
+		Meta:          continuity.Meta{ID: handoffID(s), Reason: handoffReason(s)},
+		Anchor:        continuity.Anchor{GitSHA: sha, Branch: strings.TrimSpace(string(branchOut)), Dirty: dirty},
+		Action:        authored.Action,
+		Command:       command,
+		Remaining:     authored.Remaining,
+		DeadEnds:      authored.DeadEnds,
+		OpenQuestions: authored.OpenQuestions,
+		Learned:       authored.Learned,
 	}, nil
 }

+type handoffAnswer struct {
+	Action, Why            string
+	Remaining              []string
+	DeadEnds               []continuity.DeadEnd
+	OpenQuestions, Learned []string
+}
+
+func parseHandoffAnswer(answer string) (handoffAnswer, error) {
+	var out handoffAnswer
+	sections := map[string][]string{}
+	var current string
+	for _, raw := range strings.Split(strings.ReplaceAll(answer, "\r\n", "\n"), "\n") {
+		line := strings.TrimSpace(raw)
+		if line == "" {
+			continue
+		}
+		for _, name := range []string{"NEXT", "WHY", "REMAINING", "DEAD ENDS", "OPEN Q", "LEARNED"} {
+			prefix := name + ":"
+			if strings.HasPrefix(line, prefix) {
+				current = name
+				if value := strings.TrimSpace(strings.TrimPrefix(line, prefix)); value != "" {
+					sections[name] = append(sections[name], value)
+				}
+				goto parsed
+			}
+		}
+		if current == "" {
+			return out, fmt.Errorf("unexpected line %q", line)
+		}
+		sections[current] = append(sections[current], strings.TrimSpace(strings.TrimPrefix(line, "- ")))
+	parsed:
+	}
+	for _, name := range []string{"NEXT", "WHY", "REMAINING", "DEAD ENDS", "OPEN Q", "LEARNED"} {
+		if len(sections[name]) == 0 {
+			return out, fmt.Errorf("missing %s", name)
+		}
+	}
+	if len(sections["NEXT"]) != 1 || len(sections["WHY"]) != 1 {
+		return out, fmt.Errorf("NEXT and WHY each require one line")
+	}
+	out.Action = sections["NEXT"][0] + " — " + sections["WHY"][0]
+	for _, name := range []string{"REMAINING", "OPEN Q", "LEARNED"} {
+		values, err := answerList(sections[name])
+		if err != nil {
+			return out, fmt.Errorf("%s: %w", name, err)
+		}
+		switch name {
+		case "REMAINING":
+			out.Remaining = values
+		case "OPEN Q":
+			out.OpenQuestions = values
+		case "LEARNED":
+			out.Learned = values
+		}
+	}
+	deadEnds, err := answerList(sections["DEAD ENDS"])
+	if err != nil {
+		return out, fmt.Errorf("DEAD ENDS: %w", err)
+	}
+	for _, item := range deadEnds {
+		parts := strings.SplitN(item, "→", 2)
+		if len(parts) != 2 {
+			return out, fmt.Errorf("DEAD ENDS: want 'tried X → failed because Y'")
+		}
+		if !strings.HasPrefix(parts[0], "tried ") || !strings.HasPrefix(strings.TrimSpace(parts[1]), "failed because ") {
+			return out, fmt.Errorf("DEAD ENDS: want 'tried X → failed because Y'")
+		}
+		tried := strings.TrimSpace(strings.TrimPrefix(parts[0], "tried "))
+		why := strings.TrimSpace(strings.TrimPrefix(parts[1], "failed because "))
+		if tried == "" || why == "" {
+			return out, fmt.Errorf("DEAD ENDS: want 'tried X → failed because Y'")
+		}
+		out.DeadEnds = append(out.DeadEnds, continuity.DeadEnd{Tried: tried, WhyFailed: why})
+	}
+	if err := (continuity.Handoff{Meta: continuity.Meta{ID: "answer", Reason: "manual"}, Anchor: continuity.Anchor{GitSHA: strings.Repeat("0", 40), Branch: "answer"}, Action: out.Action, Remaining: out.Remaining, DeadEnds: out.DeadEnds, OpenQuestions: out.OpenQuestions, Learned: out.Learned}).Validate(); err != nil {
+		return out, err
+	}
+	return out, nil
+}
+
+func answerList(lines []string) ([]string, error) {
+	if len(lines) == 1 && lines[0] == "NONE" {
+		return nil, nil
+	}
+	for _, line := range lines {
+		if line == "NONE" {
+			return nil, fmt.Errorf("NONE must be the only value")
+		}
+	}
+	return lines, nil
+}
+
+func (a CLIAdapter) lastObservedCommand(s Session) string {
+	calls, err := a.Activity(context.Background(), s)
+	if err != nil {
+		return ""
+	}
+	for i := len(calls) - 1; i >= 0; i-- {
+		if calls[i].Kind == "command" {
+			return calls[i].Key
+		}
+	}
+	return ""
+}
+
+func handoffReason(s Session) string {
+	switch s.HandoffReason {
+	case "threshold", "milestone", "thrash", "manual":
+		return s.HandoffReason
+	default:
+		return "threshold"
+	}
+}
+
 func handoffID(s Session) string {
 	id := s.AgentName
 	if id == "" {
@@ -368,11 +519,12 @@ func dirtyFiles(root string) ([]continuity.Dirty, error) {
 	return dirty, nil
 }

-func agentForSession(s Session, fallback string) string {
-	if s.AgentName != "" {
-		return s.AgentName
-	}
-	return fallback // compatibility with session records created before B16
+func agentForSession(_ Session, fallback string) string {
+	// pane.release_agent identifies the harness binding, not herdr's
+	// machine-global terminal name. AgentName is only for prompt routing;
+	// passing it here is accepted by herdr but leaves the binding intact.
+	// Keep the configured harness for both new and pre-B16 session records.
+	return fallback
 }
 func (a CLIAdapter) Kill(ctx context.Context, s Session) error {
 	return a.Client.Call(ctx, "pane.close", map[string]any{"pane_id": s.PaneID}, nil)