go-sip/internal/agent/recording_delivery.go

package agent

import (
	"bytes"
	"context"
	"crypto/sha256"
	"encoding/binary"
	"encoding/hex"
	"encoding/json"
	"errors"
	"fmt"
	"strings"
	"sync"

	agentpb "git.ipao.vip/rogee/go-sip/gen/agent"
)

var ErrRecordingRetryManaged = errors.New("recording delivery already attempted; retry the original result or persisted recovery instead")

// CompletedRecording describes one confirmed call and, when expected, the
// bounded in-memory WAV created from its actual media. Recording IDs are stable
// across retries; ResultPayload has the approved call identity and recording:{}.
type CompletedRecording struct {
	ResultPayload []byte
	Expected      bool
	RecordingID   string
	UploadID      string
	WAV           []byte
	DurationMS    int64
	CaptureError  error
}

// RecordingDelivery owns one approved call. In-process retries may report a
// confirmed PUT again, but can never repeat that PUT. Failed or uncertain PUTs
// are handed to RecordingRecovery, not retried through Complete.
type RecordingDelivery struct {
	Call     RecordingClient
	Recovery *RecordingRecovery

	mu           sync.Mutex
	ended        bool
	attempted    bool
	pending      []byte
	pendingProof *agentpb.UploadObservation
}

func (d *RecordingDelivery) Complete(ctx context.Context, recording CompletedRecording) error {
	d.mu.Lock()
	defer d.mu.Unlock()
	if d.attempted {
		return ErrRecordingRetryManaged
	}
	// Release the confirmed call independently of OSS and final-result delivery.
	if !d.ended {
		if err := d.Call.ReportEnded(ctx); err != nil {
			return err
		}
		d.ended = true
	}
	result, err := emptyRecordingResult(recording.ResultPayload)
	if err != nil {
		return err
	}
	if !recording.Expected {
		if len(recording.WAV) != 0 || recording.CaptureError != nil || recording.RecordingID != "" || recording.UploadID != "" || recording.DurationMS != 0 {
			return errors.New("call without a recording cannot contain audio or upload identity")
		}
		return d.report(ctx, recording.ResultPayload, nil)
	}
	if recording.CaptureError != nil || len(recording.WAV) == 0 {
		if len(recording.WAV) != 0 {
			return errors.New("recording has both an error and apparently uploadable audio")
		}
		var reason string
		if err := json.Unmarshal(result["reason_message"], &reason); err != nil {
			return errors.New("failed recording needs the original call reason")
		}
		if reason != "" {
			reason += "; "
		}
		reason += "recording generation failed"
		result["reason_message"], err = json.Marshal(reason)
		if err != nil {
			return err
		}
		payload, err := json.Marshal(result)
		if err != nil {
			return err
		}
		return d.report(ctx, payload, nil)
	}
	if recording.RecordingID == "" || recording.UploadID == "" || !validMonoWAV(recording.WAV, recording.DurationMS) {
		return errors.New("recording identity or 16-kHz mono WAV is invalid")
	}
	if d.Recovery == nil || strings.TrimSpace(d.Recovery.Root) == "" {
		return errors.New("private recording recovery directory and upload client are required before PUT")
	}
	sum := sha256.Sum256(recording.WAV)
	digest := hex.EncodeToString(sum[:])
	asset := &agentpb.AssetDescriptor{
		Kind: agentpb.AssetKind_ASSET_KIND_RECORDING, AssetId: recording.RecordingID,
		CallId: d.Call.SourceEventID, ExecutionId: d.Call.SourceEventID,
		Format: "wav", Channels: 1, SampleRateHz: 16000, DurationMs: recording.DurationMS,
		SizeBytes: int64(len(recording.WAV)), ChecksumSha256: digest,
	}
	grant, err := d.Call.RequestUpload(ctx, asset, recording.UploadID)
	if err != nil {
		return err
	}
	result["recording"], err = json.Marshal(struct {
		Status         string `json:"status"`
		Bucket         string `json:"bucket"`
		ObjectKey      string `json:"object_key"`
		Format         string `json:"format"`
		Channels       int    `json:"channels"`
		SampleRateHz   int    `json:"sample_rate_hz"`
		DurationMS     int64  `json:"duration_ms"`
		SizeBytes      int64  `json:"size_bytes"`
		ChecksumSHA256 string `json:"checksum_sha256"`
	}{"uploaded", grant.Bucket, grant.ObjectKey, "wav", 1, 16000, recording.DurationMS, int64(len(recording.WAV)), digest})
	if err != nil {
		return err
	}
	payload, err := json.Marshal(result)
	if err != nil {
		return err
	}

	// A returned OSS failure needs the original exact result and bytes; no
	// failure in this path is allowed to fabricate an uploaded result.
	d.attempted = true
	uploaded, err := d.Recovery.Upload.UploadBytes(ctx, grant, recording.WAV)
	if err != nil {
		var rejected *UploadHTTPError
		if errors.As(err, &rejected) || errors.Is(err, ErrUploadOutcomeUnknown) {
			entry := RecordingRecoveryEntry{
				CallID: d.Call.SourceEventID, SourceEventID: d.Call.SourceEventID,
				RecordingID: recording.RecordingID, UploadID: recording.UploadID,
				Bucket: grant.Bucket, ObjectKey: grant.ObjectKey,
				SHA256: digest, SizeBytes: int64(len(recording.WAV)), ResultPayload: payload,
			}
			if _, saveErr := d.Recovery.SaveFailure(recording.WAV, entry, err); saveErr != nil {
				return errors.Join(err, fmt.Errorf("persist failed recording for recovery: %w", saveErr))
			}
		}
		return err
	}
	proof := &agentpb.UploadObservation{
		UploadId: recording.UploadID, RecordingId: recording.RecordingID,
		PutStatusCode: int32(uploaded.StatusCode), SizeBytes: uploaded.SizeBytes, ChecksumSha256: uploaded.SHA256,
	}
	return d.report(ctx, payload, proof)
}

// RetryResult reports only the original result after a confirmed PUT (or a
// no-recording result); it never requests another upload or sends audio.
func (d *RecordingDelivery) RetryResult(ctx context.Context) error {
	d.mu.Lock()
	defer d.mu.Unlock()
	if len(d.pending) == 0 {
		return ErrRecordingRetryManaged
	}
	return d.report(ctx, d.pending, d.pendingProof)
}

func (d *RecordingDelivery) report(ctx context.Context, payload []byte, proof *agentpb.UploadObservation) error {
	d.attempted = true
	d.pending = bytes.Clone(payload)
	d.pendingProof = proof
	if _, err := d.Call.ReportFinal(ctx, d.pending, d.pendingProof); err != nil {
		return err
	}
	d.pending = nil
	d.pendingProof = nil
	return nil
}

func emptyRecordingResult(payload []byte) (map[string]json.RawMessage, error) {
	var result map[string]json.RawMessage
	if err := json.Unmarshal(payload, &result); err != nil || result == nil {
		return nil, errors.New("final call result must be a JSON object")
	}
	var recording map[string]json.RawMessage
	if err := json.Unmarshal(result["recording"], &recording); err != nil || recording == nil || len(recording) != 0 {
		return nil, errors.New("final call result must start with an empty recording object")
	}
	return result, nil
}

func validMonoWAV(wav []byte, durationMS int64) bool {
	if len(wav) <= 44 || len(wav)%2 != 0 || uint64(len(wav)) > uint64(^uint32(0))+8 || durationMS < 0 {
		return false
	}
	return string(wav[:4]) == "RIFF" && binary.LittleEndian.Uint32(wav[4:8]) == uint32(len(wav)-8) &&
		string(wav[8:16]) == "WAVEfmt " && binary.LittleEndian.Uint32(wav[16:20]) == 16 &&
		binary.LittleEndian.Uint16(wav[20:22]) == 1 && binary.LittleEndian.Uint16(wav[22:24]) == 1 &&
		binary.LittleEndian.Uint32(wav[24:28]) == 16000 && binary.LittleEndian.Uint32(wav[28:32]) == 32000 &&
		binary.LittleEndian.Uint16(wav[32:34]) == 2 && binary.LittleEndian.Uint16(wav[34:36]) == 16 &&
		string(wav[36:40]) == "data" && binary.LittleEndian.Uint32(wav[40:44]) == uint32(len(wav)-44) &&
		int64(len(wav)-44)*1000/32000 == durationMS
}