Files
multica/server/internal/handler/agent.go
Bohan Jiang ce447c7f06 feat(agent): add custom CLI arguments support (#986)
* feat(agent): add custom CLI arguments support

Allow users to configure custom CLI arguments per agent that get
appended to the agent subprocess command at launch time. This enables
use cases like specifying different models (--model o3), max turns,
or other provider-specific flags without needing separate runtimes.

Changes:
- Add custom_args JSONB column to agent table (migration 041)
- Update API handler to accept/return custom_args in create/update
- Pass custom_args through claim endpoint to daemon
- Append custom_args to CLI commands for all agent backends
- Add ExecOptions.CustomArgs field in agent package
- Add Custom Args tab in agent detail UI
- Add --custom-args flag to CLI agent create/update commands

Closes MUL-802

* fix(agent): filter protocol-critical flags from custom_args

Add per-backend filtering of custom_args to prevent users from
accidentally overriding flags that the daemon hardcodes for its
communication protocol (e.g. --output-format, --input-format,
--permission-mode for Claude).

This follows the same pattern as custom_env's isBlockedEnvKey: we
only block the small, stable set of flags that would break the
daemon↔agent protocol — not every possible dangerous flag. Workspace
members are trusted for everything else.

Each backend defines its own blocked set:
- Claude: -p, --output-format, --input-format, --permission-mode
- Gemini: -p, --yolo, -o
- Codex: --listen
- OpenCode: --format
- OpenClaw: --local, --json, --session-id, --message
- Hermes: none (ACP is positional)

Includes unit tests for the filtering logic.

* fix(agent): address code review nits for custom_args

- Replace module-level `nextArgId` counter with `crypto.randomUUID()`
  in custom-args-tab.tsx to avoid SSR ID conflicts
- Add unit tests for custom args passthrough and blocked-arg filtering
  in both Claude and Gemini arg builders
2026-04-15 14:58:53 +08:00

562 lines
19 KiB
Go

package handler
import (
"encoding/json"
"log/slog"
"net/http"
"github.com/go-chi/chi/v5"
"github.com/jackc/pgx/v5/pgtype"
"github.com/multica-ai/multica/server/internal/logger"
"github.com/multica-ai/multica/server/internal/service"
db "github.com/multica-ai/multica/server/pkg/db/generated"
"github.com/multica-ai/multica/server/pkg/protocol"
)
type AgentResponse struct {
ID string `json:"id"`
WorkspaceID string `json:"workspace_id"`
RuntimeID string `json:"runtime_id"`
Name string `json:"name"`
Description string `json:"description"`
Instructions string `json:"instructions"`
AvatarURL *string `json:"avatar_url"`
RuntimeMode string `json:"runtime_mode"`
RuntimeConfig any `json:"runtime_config"`
CustomEnv map[string]string `json:"custom_env"`
CustomArgs []string `json:"custom_args"`
CustomEnvRedacted bool `json:"custom_env_redacted"`
Visibility string `json:"visibility"`
Status string `json:"status"`
MaxConcurrentTasks int32 `json:"max_concurrent_tasks"`
OwnerID *string `json:"owner_id"`
Skills []SkillResponse `json:"skills"`
CreatedAt string `json:"created_at"`
UpdatedAt string `json:"updated_at"`
ArchivedAt *string `json:"archived_at"`
ArchivedBy *string `json:"archived_by"`
}
func agentToResponse(a db.Agent) AgentResponse {
var rc any
if a.RuntimeConfig != nil {
json.Unmarshal(a.RuntimeConfig, &rc)
}
if rc == nil {
rc = map[string]any{}
}
var customEnv map[string]string
if a.CustomEnv != nil {
if err := json.Unmarshal(a.CustomEnv, &customEnv); err != nil {
slog.Warn("failed to unmarshal agent custom_env", "agent_id", uuidToString(a.ID), "error", err)
}
}
if customEnv == nil {
customEnv = map[string]string{}
}
var customArgs []string
if a.CustomArgs != nil {
if err := json.Unmarshal(a.CustomArgs, &customArgs); err != nil {
slog.Warn("failed to unmarshal agent custom_args", "agent_id", uuidToString(a.ID), "error", err)
}
}
if customArgs == nil {
customArgs = []string{}
}
return AgentResponse{
ID: uuidToString(a.ID),
WorkspaceID: uuidToString(a.WorkspaceID),
RuntimeID: uuidToString(a.RuntimeID),
Name: a.Name,
Description: a.Description,
Instructions: a.Instructions,
AvatarURL: textToPtr(a.AvatarUrl),
RuntimeMode: a.RuntimeMode,
RuntimeConfig: rc,
CustomEnv: customEnv,
CustomArgs: customArgs,
Visibility: a.Visibility,
Status: a.Status,
MaxConcurrentTasks: a.MaxConcurrentTasks,
OwnerID: uuidToPtr(a.OwnerID),
Skills: []SkillResponse{},
CreatedAt: timestampToString(a.CreatedAt),
UpdatedAt: timestampToString(a.UpdatedAt),
ArchivedAt: timestampToPtr(a.ArchivedAt),
ArchivedBy: uuidToPtr(a.ArchivedBy),
}
}
// RepoData holds repository information included in claim responses so the
// daemon can set up worktrees for each workspace repo.
type RepoData struct {
URL string `json:"url"`
Description string `json:"description"`
}
type AgentTaskResponse struct {
ID string `json:"id"`
AgentID string `json:"agent_id"`
RuntimeID string `json:"runtime_id"`
IssueID string `json:"issue_id"`
WorkspaceID string `json:"workspace_id"`
Status string `json:"status"`
Priority int32 `json:"priority"`
DispatchedAt *string `json:"dispatched_at"`
StartedAt *string `json:"started_at"`
CompletedAt *string `json:"completed_at"`
Result any `json:"result"`
Error *string `json:"error"`
Agent *TaskAgentData `json:"agent,omitempty"`
Repos []RepoData `json:"repos,omitempty"`
CreatedAt string `json:"created_at"`
PriorSessionID string `json:"prior_session_id,omitempty"` // session ID from a previous task on same issue
PriorWorkDir string `json:"prior_work_dir,omitempty"` // work_dir from a previous task on same issue
TriggerCommentID *string `json:"trigger_comment_id,omitempty"` // comment that triggered this task
TriggerCommentContent string `json:"trigger_comment_content,omitempty"` // content of the triggering comment
ChatSessionID string `json:"chat_session_id,omitempty"` // non-empty for chat tasks
ChatMessage string `json:"chat_message,omitempty"` // user message for chat tasks
}
// TaskAgentData holds agent info included in claim responses so the daemon
// can set up the execution environment (branch naming, skill files, instructions).
type TaskAgentData struct {
ID string `json:"id"`
Name string `json:"name"`
Instructions string `json:"instructions"`
Skills []service.AgentSkillData `json:"skills,omitempty"`
CustomEnv map[string]string `json:"custom_env,omitempty"`
CustomArgs []string `json:"custom_args,omitempty"`
}
func taskToResponse(t db.AgentTaskQueue) AgentTaskResponse {
var result any
if t.Result != nil {
json.Unmarshal(t.Result, &result)
}
return AgentTaskResponse{
ID: uuidToString(t.ID),
AgentID: uuidToString(t.AgentID),
RuntimeID: uuidToString(t.RuntimeID),
IssueID: uuidToString(t.IssueID),
Status: t.Status,
Priority: t.Priority,
DispatchedAt: timestampToPtr(t.DispatchedAt),
StartedAt: timestampToPtr(t.StartedAt),
CompletedAt: timestampToPtr(t.CompletedAt),
Result: result,
Error: textToPtr(t.Error),
CreatedAt: timestampToString(t.CreatedAt),
TriggerCommentID: uuidToPtr(t.TriggerCommentID),
}
}
func (h *Handler) ListAgents(w http.ResponseWriter, r *http.Request) {
workspaceID := resolveWorkspaceID(r)
member, ok := h.workspaceMember(w, r, workspaceID)
if !ok {
return
}
userID := requestUserID(r)
var agents []db.Agent
var err error
if r.URL.Query().Get("include_archived") == "true" {
agents, err = h.Queries.ListAllAgents(r.Context(), parseUUID(workspaceID))
} else {
agents, err = h.Queries.ListAgents(r.Context(), parseUUID(workspaceID))
}
if err != nil {
writeError(w, http.StatusInternalServerError, "failed to list agents")
return
}
// Batch-load skills for all agents to avoid N+1.
skillRows, err := h.Queries.ListAgentSkillsByWorkspace(r.Context(), parseUUID(workspaceID))
if err != nil {
writeError(w, http.StatusInternalServerError, "failed to load agent skills")
return
}
skillMap := map[string][]SkillResponse{}
for _, row := range skillRows {
agentID := uuidToString(row.AgentID)
skillMap[agentID] = append(skillMap[agentID], SkillResponse{
ID: uuidToString(row.ID),
Name: row.Name,
Description: row.Description,
})
}
// All agents (including private) are visible to workspace members.
visible := make([]AgentResponse, 0, len(agents))
for _, a := range agents {
resp := agentToResponse(a)
if skills, ok := skillMap[resp.ID]; ok {
resp.Skills = skills
}
// Redact custom_env for users who are not the agent owner or workspace owner/admin.
if !canViewAgentEnv(a, userID, member.Role) {
redactEnv(&resp)
}
visible = append(visible, resp)
}
writeJSON(w, http.StatusOK, visible)
}
func (h *Handler) GetAgent(w http.ResponseWriter, r *http.Request) {
id := chi.URLParam(r, "id")
agent, ok := h.loadAgentForUser(w, r, id)
if !ok {
return
}
resp := agentToResponse(agent)
skills, err := h.Queries.ListAgentSkills(r.Context(), agent.ID)
if err != nil {
writeError(w, http.StatusInternalServerError, "failed to load agent skills")
return
}
if len(skills) > 0 {
resp.Skills = make([]SkillResponse, len(skills))
for i, s := range skills {
resp.Skills[i] = skillToResponse(s)
}
}
// Redact custom_env for users who are not the agent owner or workspace owner/admin.
userID := requestUserID(r)
if member, ok := ctxMember(r.Context()); ok {
if !canViewAgentEnv(agent, userID, member.Role) {
redactEnv(&resp)
}
}
writeJSON(w, http.StatusOK, resp)
}
type CreateAgentRequest struct {
Name string `json:"name"`
Description string `json:"description"`
Instructions string `json:"instructions"`
AvatarURL *string `json:"avatar_url"`
RuntimeID string `json:"runtime_id"`
RuntimeConfig any `json:"runtime_config"`
CustomEnv map[string]string `json:"custom_env"`
CustomArgs []string `json:"custom_args"`
Visibility string `json:"visibility"`
MaxConcurrentTasks int32 `json:"max_concurrent_tasks"`
}
func (h *Handler) CreateAgent(w http.ResponseWriter, r *http.Request) {
workspaceID := resolveWorkspaceID(r)
var req CreateAgentRequest
if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
writeError(w, http.StatusBadRequest, "invalid request body")
return
}
ownerID, ok := requireUserID(w, r)
if !ok {
return
}
if req.Name == "" {
writeError(w, http.StatusBadRequest, "name is required")
return
}
if req.RuntimeID == "" {
writeError(w, http.StatusBadRequest, "runtime_id is required")
return
}
if req.Visibility == "" {
req.Visibility = "private"
}
if req.MaxConcurrentTasks == 0 {
req.MaxConcurrentTasks = 6
}
runtime, err := h.Queries.GetAgentRuntimeForWorkspace(r.Context(), db.GetAgentRuntimeForWorkspaceParams{
ID: parseUUID(req.RuntimeID),
WorkspaceID: parseUUID(workspaceID),
})
if err != nil {
writeError(w, http.StatusBadRequest, "invalid runtime_id")
return
}
rc, _ := json.Marshal(req.RuntimeConfig)
if req.RuntimeConfig == nil {
rc = []byte("{}")
}
ce, _ := json.Marshal(req.CustomEnv)
if req.CustomEnv == nil {
ce = []byte("{}")
}
ca, _ := json.Marshal(req.CustomArgs)
if req.CustomArgs == nil {
ca = []byte("[]")
}
agent, err := h.Queries.CreateAgent(r.Context(), db.CreateAgentParams{
WorkspaceID: parseUUID(workspaceID),
Name: req.Name,
Description: req.Description,
Instructions: req.Instructions,
AvatarUrl: ptrToText(req.AvatarURL),
RuntimeMode: runtime.RuntimeMode,
RuntimeConfig: rc,
RuntimeID: runtime.ID,
Visibility: req.Visibility,
MaxConcurrentTasks: req.MaxConcurrentTasks,
OwnerID: parseUUID(ownerID),
CustomEnv: ce,
CustomArgs: ca,
})
if err != nil {
slog.Warn("create agent failed", append(logger.RequestAttrs(r), "error", err, "workspace_id", workspaceID)...)
writeError(w, http.StatusInternalServerError, "failed to create agent: "+err.Error())
return
}
slog.Info("agent created", append(logger.RequestAttrs(r), "agent_id", uuidToString(agent.ID), "name", agent.Name, "workspace_id", workspaceID)...)
if runtime.Status == "online" {
h.TaskService.ReconcileAgentStatus(r.Context(), agent.ID)
agent, _ = h.Queries.GetAgent(r.Context(), agent.ID)
}
resp := agentToResponse(agent)
actorType, actorID := h.resolveActor(r, ownerID, workspaceID)
h.publish(protocol.EventAgentCreated, workspaceID, actorType, actorID, map[string]any{"agent": resp})
writeJSON(w, http.StatusCreated, resp)
}
type UpdateAgentRequest struct {
Name *string `json:"name"`
Description *string `json:"description"`
Instructions *string `json:"instructions"`
AvatarURL *string `json:"avatar_url"`
RuntimeID *string `json:"runtime_id"`
RuntimeConfig any `json:"runtime_config"`
CustomEnv *map[string]string `json:"custom_env"`
CustomArgs *[]string `json:"custom_args"`
Visibility *string `json:"visibility"`
Status *string `json:"status"`
MaxConcurrentTasks *int32 `json:"max_concurrent_tasks"`
}
// canViewAgentEnv checks whether the requesting user is allowed to see the
// agent's custom environment variables. Only the agent owner or workspace
// owner/admin may view them; for everyone else the field is redacted.
func canViewAgentEnv(agent db.Agent, userID string, memberRole string) bool {
if roleAllowed(memberRole, "owner", "admin") {
return true
}
return uuidToString(agent.OwnerID) == userID
}
// redactEnv masks custom_env values in the response when the caller is not
// authorised to view them. Keys are preserved so members can see which
// variables are configured; values are replaced with "****".
func redactEnv(resp *AgentResponse) {
masked := make(map[string]string, len(resp.CustomEnv))
for k := range resp.CustomEnv {
masked[k] = "****"
}
resp.CustomEnv = masked
resp.CustomEnvRedacted = true
}
// canManageAgent checks whether the current user can update or archive an agent.
// Only the agent owner or workspace owner/admin can manage any agent,
// regardless of whether it is public or private.
func (h *Handler) canManageAgent(w http.ResponseWriter, r *http.Request, agent db.Agent) bool {
wsID := uuidToString(agent.WorkspaceID)
member, ok := h.requireWorkspaceRole(w, r, wsID, "agent not found", "owner", "admin", "member")
if !ok {
return false
}
isAdmin := roleAllowed(member.Role, "owner", "admin")
isAgentOwner := uuidToString(agent.OwnerID) == requestUserID(r)
if !isAdmin && !isAgentOwner {
writeError(w, http.StatusForbidden, "only the agent owner can manage this agent")
return false
}
return true
}
func (h *Handler) UpdateAgent(w http.ResponseWriter, r *http.Request) {
id := chi.URLParam(r, "id")
agent, ok := h.loadAgentForUser(w, r, id)
if !ok {
return
}
if !h.canManageAgent(w, r, agent) {
return
}
var req UpdateAgentRequest
if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
writeError(w, http.StatusBadRequest, "invalid request body")
return
}
params := db.UpdateAgentParams{
ID: parseUUID(id),
}
if req.Name != nil {
params.Name = pgtype.Text{String: *req.Name, Valid: true}
}
if req.Description != nil {
params.Description = pgtype.Text{String: *req.Description, Valid: true}
}
if req.Instructions != nil {
params.Instructions = pgtype.Text{String: *req.Instructions, Valid: true}
}
if req.AvatarURL != nil {
params.AvatarUrl = pgtype.Text{String: *req.AvatarURL, Valid: true}
}
if req.RuntimeConfig != nil {
rc, _ := json.Marshal(req.RuntimeConfig)
params.RuntimeConfig = rc
}
if req.CustomEnv != nil {
ce, _ := json.Marshal(*req.CustomEnv)
params.CustomEnv = ce
}
if req.CustomArgs != nil {
ca, _ := json.Marshal(*req.CustomArgs)
params.CustomArgs = ca
}
if req.RuntimeID != nil {
runtime, err := h.Queries.GetAgentRuntimeForWorkspace(r.Context(), db.GetAgentRuntimeForWorkspaceParams{
ID: parseUUID(*req.RuntimeID),
WorkspaceID: agent.WorkspaceID,
})
if err != nil {
writeError(w, http.StatusBadRequest, "invalid runtime_id")
return
}
params.RuntimeID = runtime.ID
params.RuntimeMode = pgtype.Text{String: runtime.RuntimeMode, Valid: true}
}
if req.Visibility != nil {
params.Visibility = pgtype.Text{String: *req.Visibility, Valid: true}
}
if req.Status != nil {
params.Status = pgtype.Text{String: *req.Status, Valid: true}
}
if req.MaxConcurrentTasks != nil {
params.MaxConcurrentTasks = pgtype.Int4{Int32: *req.MaxConcurrentTasks, Valid: true}
}
agent, err := h.Queries.UpdateAgent(r.Context(), params)
if err != nil {
slog.Warn("update agent failed", append(logger.RequestAttrs(r), "error", err, "agent_id", id)...)
writeError(w, http.StatusInternalServerError, "failed to update agent: "+err.Error())
return
}
resp := agentToResponse(agent)
slog.Info("agent updated", append(logger.RequestAttrs(r), "agent_id", id, "workspace_id", uuidToString(agent.WorkspaceID))...)
userID := requestUserID(r)
actorType, actorID := h.resolveActor(r, userID, uuidToString(agent.WorkspaceID))
h.publish(protocol.EventAgentStatus, uuidToString(agent.WorkspaceID), actorType, actorID, map[string]any{"agent": resp})
writeJSON(w, http.StatusOK, resp)
}
func (h *Handler) ArchiveAgent(w http.ResponseWriter, r *http.Request) {
id := chi.URLParam(r, "id")
agent, ok := h.loadAgentForUser(w, r, id)
if !ok {
return
}
if !h.canManageAgent(w, r, agent) {
return
}
if agent.ArchivedAt.Valid {
writeError(w, http.StatusConflict, "agent is already archived")
return
}
userID := requestUserID(r)
archived, err := h.Queries.ArchiveAgent(r.Context(), db.ArchiveAgentParams{
ID: parseUUID(id),
ArchivedBy: parseUUID(userID),
})
if err != nil {
slog.Warn("archive agent failed", append(logger.RequestAttrs(r), "error", err, "agent_id", id)...)
writeError(w, http.StatusInternalServerError, "failed to archive agent")
return
}
// Cancel all pending/active tasks for this agent.
if err := h.Queries.CancelAgentTasksByAgent(r.Context(), parseUUID(id)); err != nil {
slog.Warn("cancel agent tasks on archive failed", append(logger.RequestAttrs(r), "error", err, "agent_id", id)...)
}
wsID := uuidToString(archived.WorkspaceID)
slog.Info("agent archived", append(logger.RequestAttrs(r), "agent_id", id, "workspace_id", wsID)...)
resp := agentToResponse(archived)
actorType, actorID := h.resolveActor(r, userID, wsID)
h.publish(protocol.EventAgentArchived, wsID, actorType, actorID, map[string]any{"agent": resp})
writeJSON(w, http.StatusOK, resp)
}
func (h *Handler) RestoreAgent(w http.ResponseWriter, r *http.Request) {
id := chi.URLParam(r, "id")
agent, ok := h.loadAgentForUser(w, r, id)
if !ok {
return
}
if !h.canManageAgent(w, r, agent) {
return
}
if !agent.ArchivedAt.Valid {
writeError(w, http.StatusConflict, "agent is not archived")
return
}
restored, err := h.Queries.RestoreAgent(r.Context(), parseUUID(id))
if err != nil {
slog.Warn("restore agent failed", append(logger.RequestAttrs(r), "error", err, "agent_id", id)...)
writeError(w, http.StatusInternalServerError, "failed to restore agent")
return
}
wsID := uuidToString(restored.WorkspaceID)
slog.Info("agent restored", append(logger.RequestAttrs(r), "agent_id", id, "workspace_id", wsID)...)
resp := agentToResponse(restored)
userID := requestUserID(r)
actorType, actorID := h.resolveActor(r, userID, wsID)
h.publish(protocol.EventAgentRestored, wsID, actorType, actorID, map[string]any{"agent": resp})
writeJSON(w, http.StatusOK, resp)
}
func (h *Handler) ListAgentTasks(w http.ResponseWriter, r *http.Request) {
id := chi.URLParam(r, "id")
if _, ok := h.loadAgentForUser(w, r, id); !ok {
return
}
tasks, err := h.Queries.ListAgentTasks(r.Context(), parseUUID(id))
if err != nil {
writeError(w, http.StatusInternalServerError, "failed to list agent tasks")
return
}
resp := make([]AgentTaskResponse, len(tasks))
for i, t := range tasks {
resp[i] = taskToResponse(t)
}
writeJSON(w, http.StatusOK, resp)
}