Skip to content

feat: add neutral SCM observability #46

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
illegalcall merged 1 commit into from
May 31, 2026
Merged

feat: add neutral SCM observability #46

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
167 changes: 146 additions & 21 deletions backend/internal/adapters/scm/github/client.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@ import (
"encoding/json"
"errors"
"io"
"log/slog"
"net"
"net/http"
"net/url"
Expand All @@ -14,6 +15,7 @@ import (
"time"

"github.com/aoagents/agent-orchestrator/backend/internal/domain"
scmlog "github.com/aoagents/agent-orchestrator/backend/internal/scm/logging"
)

const (
Expand All @@ -28,6 +30,7 @@ type Client struct {
restBase string
graphqlURL string
userAgent string
logger *slog.Logger
}

type ClientOptions struct {
Expand All @@ -36,10 +39,11 @@ type ClientOptions struct {
RESTBase string
GraphQLURL string
UserAgent string
Logger *slog.Logger
}

func NewClient(opts ClientOptions) *Client {
c := &Client{httpClient: opts.HTTPClient, tokens: opts.Token, restBase: opts.RESTBase, graphqlURL: opts.GraphQLURL, userAgent: opts.UserAgent}
c := &Client{httpClient: opts.HTTPClient, tokens: opts.Token, restBase: opts.RESTBase, graphqlURL: opts.GraphQLURL, userAgent: opts.UserAgent, logger: opts.Logger}
if c.httpClient == nil {
c.httpClient = http.DefaultClient
}
Expand Down Expand Up @@ -76,22 +80,31 @@ type RESTResponse struct {
}

func (c *Client) DoREST(ctx context.Context, method, path string, q url.Values, body any, etag string, operation string) (RESTResponse, error) {
ctx, _ = scmlog.EnsureCorrelationID(ctx)
started := time.Now()
logger := scmlog.Logger(c.logger)
endpoint := githubEndpointTemplate(path)
var rdr io.Reader
if body != nil {
b, err := json.Marshal(body)
if err != nil {
return RESTResponse{}, &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
se := &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
logTransportFailure(ctx, logger, operation, method, endpoint, started, 0, nil, false, false, se)
return RESTResponse{}, se
}
rdr = bytes.NewReader(b)
}
u, err := c.restURL(path, q)
if err != nil {
return RESTResponse{}, &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
se := &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
logTransportFailure(ctx, logger, operation, method, endpoint, started, 0, nil, false, etag != "", se)
return RESTResponse{}, se
}
req, err := http.NewRequestWithContext(ctx, method, u, rdr)
if err != nil {
return RESTResponse{}, &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
se := &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
logTransportFailure(ctx, logger, operation, method, endpoint, started, 0, nil, false, etag != "", se)
return RESTResponse{}, se
}
if body != nil {
req.Header.Set("Content-Type", "application/json")
Expand All @@ -102,60 +115,89 @@ func (c *Client) DoREST(ctx context.Context, method, path string, q url.Values,
if etag != "" {
req.Header.Set("If-None-Match", etag)
}
logTransportRequest(ctx, logger, operation, method, endpoint, etag != "")
if err := c.authorize(ctx, req); err != nil {
logTransportFailure(ctx, logger, operation, method, endpoint, started, 0, nil, false, etag != "", err)
return RESTResponse{}, err
}

resp, err := c.httpClient.Do(req)
if err != nil {
return RESTResponse{}, normalizeHTTPError(operation, err)
se := normalizeHTTPError(operation, err)
logTransportFailure(ctx, logger, operation, method, endpoint, started, 0, nil, false, etag != "", se)
return RESTResponse{}, se
}
defer resp.Body.Close()
b, readErr := io.ReadAll(resp.Body)
if readErr != nil {
return RESTResponse{}, &domain.SCMError{Kind: domain.SCMErrorNetwork, Operation: operation, Message: readErr.Error(), Cause: readErr}
se := &domain.SCMError{Kind: domain.SCMErrorNetwork, Operation: operation, Message: readErr.Error(), Cause: readErr}
logTransportFailure(ctx, logger, operation, method, endpoint, started, resp.StatusCode, rateLimitFromHeaders(resp.Header), false, etag != "", se)
return RESTResponse{}, se
}
rl := rateLimitFromHeaders(resp.Header)
out := RESTResponse{StatusCode: resp.StatusCode, NotModified: resp.StatusCode == http.StatusNotModified, ETag: resp.Header.Get("ETag"), Body: b, RateLimit: rl, Diagnostic: domain.SCMDiagnostic{Operation: operation, StatusCode: resp.StatusCode, ETag: resp.Header.Get("ETag"), CacheHit: resp.StatusCode == http.StatusNotModified, StartedAt: started, DurationMS: time.Since(started).Milliseconds()}}
if resp.StatusCode == http.StatusNotModified {
logTransportResponse(ctx, logger, operation, method, endpoint, started, resp.StatusCode, rl, out.NotModified, out.ETag != "")
return out, nil
}
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
return out, githubStatusError(operation, resp.StatusCode, b, rl)
se := githubStatusError(operation, resp.StatusCode, b, rl)
out.Diagnostic.ErrorKind = se.Kind
out.Diagnostic.Message = scmlog.SafeDiagnosticMessage(se)
logTransportFailure(ctx, logger, operation, method, endpoint, started, resp.StatusCode, rl, false, out.ETag != "" || etag != "", se)
return out, se
}
logTransportResponse(ctx, logger, operation, method, endpoint, started, resp.StatusCode, rl, out.NotModified, out.ETag != "")
return out, nil
}

func (c *Client) DoGraphQL(ctx context.Context, query string, variables map[string]any, operation string) (map[string]any, *domain.SCMRateLimit, domain.SCMDiagnostic, error) {
ctx, _ = scmlog.EnsureCorrelationID(ctx)
started := time.Now()
logger := scmlog.Logger(c.logger)
const endpoint = "/graphql"
body := map[string]any{"query": query, "variables": variables}
b, err := json.Marshal(body)
if err != nil {
return nil, nil, domain.SCMDiagnostic{}, &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
se := &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
logTransportFailure(ctx, logger, operation, http.MethodPost, endpoint, started, 0, nil, false, false, se)
return nil, nil, domain.SCMDiagnostic{}, se
}
req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.graphqlURL, bytes.NewReader(b))
if err != nil {
return nil, nil, domain.SCMDiagnostic{}, &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
se := &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
logTransportFailure(ctx, logger, operation, http.MethodPost, endpoint, started, 0, nil, false, false, se)
return nil, nil, domain.SCMDiagnostic{}, se
}
req.Header.Set("Content-Type", "application/json")
req.Header.Set("Accept", "application/json")
req.Header.Set("User-Agent", c.userAgent)
logTransportRequest(ctx, logger, operation, http.MethodPost, endpoint, false)
if err := c.authorize(ctx, req); err != nil {
logTransportFailure(ctx, logger, operation, http.MethodPost, endpoint, started, 0, nil, false, false, err)
return nil, nil, domain.SCMDiagnostic{}, err
}
resp, err := c.httpClient.Do(req)
if err != nil {
return nil, nil, domain.SCMDiagnostic{}, normalizeHTTPError(operation, err)
se := normalizeHTTPError(operation, err)
logTransportFailure(ctx, logger, operation, http.MethodPost, endpoint, started, 0, nil, false, false, se)
return nil, nil, domain.SCMDiagnostic{}, se
}
defer resp.Body.Close()
respBody, readErr := io.ReadAll(resp.Body)
if readErr != nil {
return nil, nil, domain.SCMDiagnostic{}, &domain.SCMError{Kind: domain.SCMErrorNetwork, Operation: operation, Message: readErr.Error(), Cause: readErr}
se := &domain.SCMError{Kind: domain.SCMErrorNetwork, Operation: operation, Message: readErr.Error(), Cause: readErr}
logTransportFailure(ctx, logger, operation, http.MethodPost, endpoint, started, resp.StatusCode, rateLimitFromHeaders(resp.Header), false, false, se)
return nil, nil, domain.SCMDiagnostic{}, se
}
rl := rateLimitFromHeaders(resp.Header)
diag := domain.SCMDiagnostic{Operation: operation, StatusCode: resp.StatusCode, StartedAt: started, DurationMS: time.Since(started).Milliseconds()}
if resp.StatusCode < 200 || resp.StatusCode >= 300 {
return nil, rl, diag, githubStatusError(operation, resp.StatusCode, respBody, rl)
se := githubStatusError(operation, resp.StatusCode, respBody, rl)
diag.ErrorKind = se.Kind
diag.Message = scmlog.SafeDiagnosticMessage(se)
logTransportFailure(ctx, logger, operation, http.MethodPost, endpoint, started, resp.StatusCode, rl, false, false, se)
return nil, rl, diag, se
}
var decoded struct {
Data map[string]any `json:"data"`
Expand All @@ -165,17 +207,28 @@ func (c *Client) DoGraphQL(ctx context.Context, query string, variables map[stri
} `json:"errors"`
}
if err := json.Unmarshal(respBody, &decoded); err != nil {
return nil, rl, diag, &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
se := &domain.SCMError{Kind: domain.SCMErrorParse, Operation: operation, Message: err.Error(), Cause: err}
diag.ErrorKind = se.Kind
diag.Message = scmlog.SafeDiagnosticMessage(se)
logTransportFailure(ctx, logger, operation, http.MethodPost, endpoint, started, resp.StatusCode, rl, false, false, se)
return nil, rl, diag, se
}
if len(decoded.Errors) > 0 {
kind := domain.SCMErrorUnavailable
msg := decoded.Errors[0].Message
if strings.Contains(strings.ToLower(msg), "rate limit") {
kind = domain.SCMErrorRateLimited
}
return decoded.Data, graphqlRateLimit(decoded.Data, rl), diag, &domain.SCMError{Kind: kind, Operation: operation, Message: msg}
}
return decoded.Data, graphqlRateLimit(decoded.Data, rl), diag, nil
rl = graphqlRateLimit(decoded.Data, rl)
se := &domain.SCMError{Kind: kind, Operation: operation, Message: scmlog.SafeDiagnosticMessage(&domain.SCMError{Kind: kind, Message: msg})}
diag.ErrorKind = se.Kind
diag.Message = scmlog.SafeDiagnosticMessage(se)
logTransportFailure(ctx, logger, operation, http.MethodPost, endpoint, started, resp.StatusCode, rl, false, false, se)
return decoded.Data, rl, diag, se
}
rl = graphqlRateLimit(decoded.Data, rl)
logTransportResponse(ctx, logger, operation, http.MethodPost, endpoint, started, resp.StatusCode, rl, false, false)
return decoded.Data, rl, diag, nil
}

func (c *Client) authorize(ctx context.Context, req *http.Request) error {
Expand Down Expand Up @@ -211,7 +264,7 @@ func normalizeHTTPError(operation string, err error) error {
return &domain.SCMError{Kind: kind, Operation: operation, Message: err.Error(), Cause: err}
}

func githubStatusError(operation string, status int, body []byte, rl *domain.SCMRateLimit) error {
func githubStatusError(operation string, status int, body []byte, rl *domain.SCMRateLimit) *domain.SCMError {
kind := domain.SCMErrorUnavailable
switch status {
case http.StatusUnauthorized, http.StatusForbidden:
Expand All @@ -225,20 +278,92 @@ func githubStatusError(operation string, status int, body []byte, rl *domain.SCM
case http.StatusTooManyRequests:
kind = domain.SCMErrorRateLimited
}
msg := strings.TrimSpace(string(body))
var gh struct {
Message string `json:"message"`
}
if json.Unmarshal(body, &gh) == nil && gh.Message != "" {
msg = gh.Message
}
_ = json.Unmarshal(body, &gh)
msg := scmlog.StatusMessage(status, body, gh.Message)
se := &domain.SCMError{Kind: kind, Operation: operation, StatusCode: status, Message: msg}
if kind == domain.SCMErrorRateLimited && rl != nil {
se.RetryAfter = rl.ResetAt
}
return se
}

func logTransportRequest(ctx context.Context, logger *slog.Logger, operation, method, endpoint string, etagPresent bool) {
attrs := transportAttrs(ctx, operation, method, endpoint,
slog.Bool(scmlog.FieldETagPresent, etagPresent),
)
logger.Debug(scmlog.EventTransportRequest, scmlog.Args(attrs)...)
}

func logTransportResponse(ctx context.Context, logger *slog.Logger, operation, method, endpoint string, started time.Time, status int, rl *domain.SCMRateLimit, cacheHit, etagPresent bool) {
attrs := transportAttrs(ctx, operation, method, endpoint,
slog.Int(scmlog.FieldStatusCode, status),
slog.Int64(scmlog.FieldDurationMS, scmlog.DurationMS(started)),
slog.Bool(scmlog.FieldCacheHit, cacheHit),
slog.Bool(scmlog.FieldETagPresent, etagPresent),
)
attrs = append(attrs, scmlog.RateLimitAttrs(rl)...)
logger.Debug(scmlog.EventTransportResponse, scmlog.Args(attrs)...)
}

func logTransportFailure(ctx context.Context, logger *slog.Logger, operation, method, endpoint string, started time.Time, status int, rl *domain.SCMRateLimit, cacheHit, etagPresent bool, err error) {
attrs := transportAttrs(ctx, operation, method, endpoint,
slog.Int64(scmlog.FieldDurationMS, scmlog.DurationMS(started)),
slog.Bool(scmlog.FieldCacheHit, cacheHit),
slog.Bool(scmlog.FieldETagPresent, etagPresent),
)
if status != 0 {
attrs = append(attrs, slog.Int(scmlog.FieldStatusCode, status))
}
attrs = append(attrs, scmlog.RateLimitAttrs(rl)...)
attrs = append(attrs, scmlog.ErrorAttrs(err)...)
if scmlog.ErrorKind(err) == domain.SCMErrorRateLimited {
logger.Warn(scmlog.EventTransportRateLimited, scmlog.Args(attrs)...)
return
}
logger.Warn(scmlog.EventTransportFailed, scmlog.Args(attrs)...)
}

func transportAttrs(ctx context.Context, operation, method, endpoint string, attrs ...slog.Attr) []slog.Attr {
base := []slog.Attr{
scmlog.CorrelationAttr(ctx),
slog.String(scmlog.FieldProvider, string(domain.SCMProviderGitHub)),
slog.String(scmlog.FieldOperation, operation),
slog.String(scmlog.FieldMethod, method),
slog.String(scmlog.FieldEndpointTemplate, endpoint),
}
return append(base, attrs...)
}

func githubEndpointTemplate(p string) string {
if p == "" {
return ""
}
parts := strings.Split(strings.Trim(p, "/"), "/")
if len(parts) >= 3 && parts[0] == "repos" {
out := []string{"repos", "{owner}", "{repo}"}
for i := 3; i < len(parts); i++ {
part := parts[i]
prev := ""
if i > 0 {
prev = parts[i-1]
}
switch {
case prev == "pulls" || prev == "issues":
out = append(out, "{number}")
case prev == "commits":
out = append(out, "{ref}")
default:
out = append(out, part)
}
}
return "/" + strings.Join(out, "/")
}
return "/" + strings.Join(parts, "/")
}

func rateLimitFromHeaders(h http.Header) *domain.SCMRateLimit {
if h == nil || h.Get("X-RateLimit-Limit") == "" {
return nil
Expand Down
4 changes: 2 additions & 2 deletions backend/internal/adapters/scm/github/commands.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -137,7 +137,7 @@ func (p *Provider) Checkout(ctx context.Context, req ports.SCMCommandRequest) (p
res.Message = "checked out with gh pr checkout"
return res, nil
} else {
return res, &domain.SCMError{Kind: domain.SCMErrorCommand, Operation: "github.command.checkout", Message: fmt.Sprintf("git checkout failed: %v; gh fallback failed: %v", err, ghErr), Cause: err}
return res, &domain.SCMError{Kind: domain.SCMErrorCommand, Operation: "github.command.checkout", Message: "checkout command failed", Cause: fmt.Errorf("git checkout failed: %w; gh fallback failed: %v", err, ghErr)}
}
}

Expand Down Expand Up @@ -187,7 +187,7 @@ func (p *Provider) normalizeCommandRequest(ctx context.Context, req ports.SCMCom

func commandResult(req ports.SCMCommandRequest, diag domain.SCMDiagnostic) ports.SCMCommandResult {
res := ports.SCMCommandResult{Provider: domain.SCMProviderGitHub, Command: req.Command, ChangeRequest: req.ChangeRequest, PerformedAt: req.Now}
if !diag.StartedAt.IsZero() || diag.Operation != "" {
if durableDiagnostic(diag) {
res.Diagnostics = []domain.SCMDiagnostic{diag}
}
return res
Expand Down
Loading
Loading