X-Git-Url: https://git.giacomolli.uk/onechat/blobdiff_plain/5b778ad04422dd456b7d673fb653f0353c47b0fe..8fb85e5647f82e029956d81be29b3a5ac3490636:/subagent_run.go

diff --git a/subagent_run.go b/subagent_run.go
index 7d3687c..0780393 100644
--- a/subagent_run.go
+++ b/subagent_run.go
@@ -5,13 +5,14 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
-	"log"
+	"log/slog"
 	"maps"
 	"os"
 	"path/filepath"
 	"slices"
 	"strings"
 	"sync"
+	"time"
 )
 
 const (
@@ -68,7 +69,7 @@ type SubagentRuns struct {
 	report    func(name, text, status string)
 	note      func(text string) error
 	fail      func(err error)
-	logger    *log.Logger
+	logger    *slog.Logger
 	notify    func(Event)
 
 	mutex   sync.Mutex
@@ -86,6 +87,7 @@ func (r *SubagentRuns) Spawn(ctx context.Context, tasks []string, effort *string
 	}
 
 	r.mutex.Lock()
+
 	closed := r.closed
 	r.mutex.Unlock()
 
@@ -136,6 +138,7 @@ func (r *SubagentRuns) Spawn(ctx context.Context, tasks []string, effort *string
 // its log with a new run, with the view of now and the effort of its note.
 func (r *SubagentRuns) Tell(ctx context.Context, name, message string) (string, error) {
 	r.mutex.Lock()
+
 	run, found := r.running[name]
 	if found {
 		defer r.mutex.Unlock()
@@ -240,7 +243,7 @@ func (r *SubagentRuns) Resume(ctx context.Context) error {
 	for _, name := range names {
 		_, err = r.Tell(ctx, name, text)
 		if err != nil {
-			r.logger.Printf("subagent %s cannot go on after the restart, and its report tells the agent: %v", name, err)
+			r.logger.Error("subagent lost at restart", "name", name, "err", err)
 			r.report(name, subagentReport(name, []string{"failed: " + err.Error()}), subagentFailed)
 		}
 	}
@@ -256,10 +259,12 @@ func (r *SubagentRuns) Resume(ctx context.Context) error {
 // Close ends every run and waits for them. No subagent starts after Close.
 func (r *SubagentRuns) Close() {
 	r.mutex.Lock()
+
 	r.closed = true
 	for _, run := range r.running {
 		run.cancel()
 	}
+
 	r.mutex.Unlock()
 
 	r.workers.Wait()
@@ -317,12 +322,13 @@ func (r *SubagentRuns) effort(subagent *Subagent) (string, error) {
 	return "", fmt.Errorf("the log of %s does not start with a subagent note", subagent.name)
 }
 
-// start runs run in the background. Its context keeps the values of ctx, but
-// only Stop and Close end it.
+// start runs run in the background and logs its start. Its context keeps the
+// values of ctx, but only Stop and Close end it.
 func (r *SubagentRuns) start(ctx context.Context, run *SubagentRun) error {
 	ctx, run.cancel = context.WithCancel(context.WithoutCancel(ctx))
 
 	r.mutex.Lock()
+
 	_, found := r.running[run.subagent.name]
 	if r.closed || found {
 		r.mutex.Unlock()
@@ -348,6 +354,8 @@ func (r *SubagentRuns) start(ctx context.Context, run *SubagentRun) error {
 	r.workers.Add(1)
 	r.mutex.Unlock()
 
+	run.started = r.subagents.clock()
+	r.logger.Info("subagent started", "name", run.subagent.name)
 	r.notify(Event{
 		Kind: "subagent",
 		Text: run.subagent.name,
@@ -370,6 +378,7 @@ func (r *SubagentRuns) load() ([]string, error) {
 	}
 
 	var file SubagentWorkFile
+
 	err = json.Unmarshal(data, &file)
 	if err != nil {
 		return nil, fmt.Errorf("decode running subagents: %w", err)
@@ -402,21 +411,25 @@ func (r *SubagentRuns) workPath() string {
 	return filepath.Join(r.agent.dir, subagentWorkFileName)
 }
 
-// run runs the chains of run and reports its end, unless Close ended it. A
-// chain follows while messages wait after the final answer of the chain before
+// run runs the chains of run and reports and logs its end with the length of
+// the run, unless Close ended it. A chain follows while messages wait after the final answer of the chain before
 // it, and no chain failed. A chain that Stop ends gives the reply "stopped".
 func (r *SubagentRuns) run(ctx context.Context, run *SubagentRun) {
 	defer r.workers.Done()
 	defer run.cancel()
 
-	var replies []string
-	var failure error
+	var (
+		replies []string
+		failure error
+	)
+
 	stopped := false
 	status := subagentFinished
 	for more := true; more; {
 		reply, err := r.chain(ctx, run)
 
 		r.mutex.Lock()
+
 		stopped = run.stopped
 		switch {
 		case err != nil && stopped:
@@ -429,13 +442,17 @@ func (r *SubagentRuns) run(ctx context.Context, run *SubagentRun) {
 
 		more = err == nil && ctx.Err() == nil && len(run.pending) > 0
 		run.pending = nil
+
 		var saveErr error
+
 		if !more {
 			delete(r.running, run.subagent.name)
+
 			if !r.closed {
 				saveErr = r.save()
 			}
 		}
+
 		r.mutex.Unlock()
 
 		if saveErr != nil {
@@ -453,13 +470,14 @@ func (r *SubagentRuns) run(ctx context.Context, run *SubagentRun) {
 		status = subagentStopped
 	}
 
+	length := r.subagents.clock().Sub(run.started)
 	switch status {
 	case subagentStopped:
-		r.logger.Printf("subagent %s is stopped by the user", run.subagent.name)
+		r.logger.Info("subagent stopped", "name", run.subagent.name, "after", length)
 	case subagentFailed:
-		r.logger.Printf("subagent %s failed, and its report tells the agent: %v", run.subagent.name, failure)
+		r.logger.Error("subagent failed", "name", run.subagent.name, "after", length, "err", failure)
 	default:
-		r.logger.Printf("subagent %s is done, and its report goes to the agent", run.subagent.name)
+		r.logger.Info("subagent done", "name", run.subagent.name, "after", length)
 	}
 
 	r.report(run.subagent.name, subagentReport(run.subagent.name, replies), status)
@@ -520,7 +538,7 @@ func (r *SubagentRuns) take(run *SubagentRun) []Message {
 
 // SubagentRun is one run of a subagent: the view that its requests get, its
 // effort, the end of its context, the messages of Tell that wait for a request,
-// and whether Stop ended it. The mutex of SubagentRuns protects the messages
+// whether Stop ended it, and the time it started. The mutex of SubagentRuns protects the messages
 // and the stop.
 type SubagentRun struct {
 	subagent *Subagent
@@ -529,6 +547,7 @@ type SubagentRun struct {
 	cancel   context.CancelFunc
 	pending  []Message
 	stopped  bool
+	started  time.Time
 }
 
 // SubagentWorkFile is the content of work.json.