Files
Enterprise-Knowledge-Base/cli/cmd/session/resume.go
T
nullkey d3c03b2f34 feat(cli)!: v0.10 reliability, agent-UX, and command-surface hardening
A hardening + finalization pass over the agent-first CLI: correctness fixes,
richer machine-readable signals, flag/naming consistency, and a symmetric
config surface. Pre-1.0, so it includes breaking renames.

Correctness:
- agent update: resolve/validate --model/--rerank-model (was storing a bogus
  name verbatim, corrupting config.model_id).
- doctor: honor WEKNORA_HOST / WEKNORA_API_KEY (headless path no longer reports
  "no host configured").
- session ask / MCP session_ask: text answer was empty on non-TTY — the agent
  stream sets Done=true on an intermediate agent_query frame before the answer,
  and AgentAccumulator treated the first Done as terminal. Terminate on the
  `complete` event (new sdk.AgentResponseTypeComplete), not a per-frame Done.
- batch exit codes: any per-item failure collapses to operation.failed (exit 1),
  including `doc upload --recursive` partial failures — a permanent per-file
  failure (e.g. a duplicate) no longer surfaces as a retryable exit 7 an agent
  would loop on; per-item typed errors stay in the envelope.

Agent-first signals & discovery:
- error.exit_code in the envelope (type + exit_code disambiguate the
  input.invalid_argument exit-2-vs-5 split in one JSON read).
- meta.hint on empty content search and on draft doc create; doc wait fails fast
  on a never-parsing draft instead of hanging to --timeout.
- retrieval-readiness is visible in the natural flow: kb status / kb check emit
  retrieval_ready, and kb create hints the fix when no embedding model is bound
  — an unconfigured KB no longer looks silently healthy.
- schema contract completeness: every leaf declares output + >=1 example
  (drift-guarded); output strings match the meta actually emitted; chunk list
  and search docs now emit meta.total_count (both previously dropped it).
- schema tolerates a quoted multi-word command label; zero-state auth — and
  `link` with no profile — point at profile setup / the headless WEKNORA_KB_ID
  path instead of looping on `auth login`.
- id-addressed reads tolerate a redundant --kb (doc view/wait, chunk list/view
  accept and ignore it, declared in schema) so a carried-over --kb doesn't
  exit 2; streaming commands warn that --jq does not apply to an NDJSON stream.
- keep JSON-always as the default; --jq hints spell out the .data path.

Consistency & gating:
- doc create: drop the deprecated --name alias (--title only; pre-1.0 break).
- chunk list --limit aligned to 1..10000; model list --limit/-L with
  has_more/total_count; api write-gates -X PUT/PATCH (exit 10); skills install
  expands a leading ~.
- docs corrected: search docs / doc list --keyword help is case-insensitive
  (server does LOWER LIKE); AGENTS.md risk-action list (no phantom kb.init; add
  model.update / kb.config.set) and batch example (failed item carries `error`);
  session resume --message id comes from `message list`, not the stream.
- auth/profile ergonomics: env credentials are now first-class — `auth token`
  prints the active WEKNORA_API_KEY / WEKNORA_TOKEN, and auth login/logout/refresh
  give an env-aware message instead of looping on "run auth login". `auth logout`
  clears credentials but keeps the profile registered (host preserved for
  re-login); deleting a profile is `profile remove`'s job (clean logout/remove
  separation, matching gh / lark).

Config surface (symmetric read/write, in-place model edits):
- kb config now returns a secret-free KBModelConfigView (was {}); `kb config`
  reads, new `kb config set` writes; `kb init` removed (misnomer).
- kb create --chat-model: retrieval-ready in one step.
- model update: edit a model in place (id preserved, references survive) —
  rotate --api-key-stdin, change base-url / display-name / etc.
- session continue-stream renamed to session resume.

Docs: AGENTS.md is the single wire-contract source; CHANGELOG slimmed; stale
kb-init / continue-stream references removed; skill wire-vocab guard extended.
AGENTS.md / weknora-shared SKILL document retrieval_ready (a KB needs an
embedding model to be searchable), that --jq does not apply to NDJSON
streams, and the env-credential-first auth path; the KB quickstart example
now creates a retrieval-ready KB.
2026-07-06 22:53:01 +08:00

191 lines
8.6 KiB
Go

// resume.go implements `weknora session resume` —
// re-attach to an SSE event buffer for an in-progress or already-completed
// assistant message under a known session_id.
//
// Server semantics:
// - Replay-from-0 + tail: every connection replays the entire stored event
// log from index 0, then tails any new events. NOT a cursor-from-disconnect
// model. Agents that already consumed events on the original stream MUST
// dedupe (by message_id + event hash, or by AssistantMessageID + content
// fingerprint) to avoid double-processing.
// - Buffer TTL:
// redis mode - 1h hardcoded (not configurable from the CLI)
// memory mode - process lifetime (server restart = data loss)
// After TTL the server returns an error which the CLI maps to
// local.sse_stream_aborted.
//
// Output shape matches `weknora chat` and `weknora session ask` NDJSON mode:
// one CLI-injected init line carrying {session_id, message_id, profile} at
// stream head, then SDK StreamResponse events verbatim. The init line lets
// agents thread the resume to the original message in their dedupe table
// without parsing the first SDK event.
package sessioncmd
import (
"context"
"github.com/spf13/cobra"
"github.com/Tencent/WeKnora/cli/internal/cmdutil"
"github.com/Tencent/WeKnora/cli/internal/iostreams"
"github.com/Tencent/WeKnora/cli/internal/output"
sdk "github.com/Tencent/WeKnora/client"
)
// resumeFields enumerates the NDJSON init-event + raw SDK event
// vocabulary surfaced for `--format json` / `--format ndjson` discovery.
var resumeFields = []string{
"session_id", "message_id",
// SDK StreamResponse fields (pass-through): id, response_type, content,
// done, knowledge_references, assistant_message_id, session_id,
// tool_calls, data
}
// ResumeOptions captures `session resume` flag/arg state.
type ResumeOptions struct {
SessionID string
MessageID string
}
// ResumeService is the narrow SDK surface this command depends on.
// *sdk.Client satisfies it; tests substitute a fake. Compile-time check
// at the bottom of this file.
type ResumeService interface {
ContinueStream(ctx context.Context, sessionID, messageID string, cb func(*sdk.StreamResponse) error) error
}
// NewCmdResume builds `weknora session resume <session-id> --message <id>`.
func NewCmdResume(f *cmdutil.Factory) *cobra.Command {
opts := &ResumeOptions{}
cmd := &cobra.Command{
Use: "resume <session-id>",
Short: "Resume an SSE event stream for an in-progress or completed session message",
Long: `Re-attach to the SSE event buffer for an assistant message under a known session.
The server replays the entire stored event log for the given (session_id, message_id)
from index 0, then tails any new events. This is NOT a cursor-from-disconnect
model: agents that already consumed events on the original stream MUST dedupe
(by message_id or event content hash) to avoid double-processing.
Buffer TTL:
- redis mode: 1h hardcoded (not configurable from the CLI)
- memory mode: process lifetime (server restart = data loss)
After TTL expiry the CLI surfaces local.sse_stream_aborted.
Typical use cases:
- Network blip mid-stream: re-attach with the same session_id + message_id
from the original 'session ask' / 'chat' init event.
- Long-running agent invocation: poll progress without blocking the original
stream.
- Post-mortem inspection: replay the full event history of a completed
message for debugging.
Output is NDJSON (matches 'chat' / 'session ask' --format json|ndjson):
one init line at head ({session_id, message_id, profile}), then raw SDK
StreamResponse events verbatim.
Note: unlike 'chat' / 'session ask', this command always emits NDJSON
regardless of --format value. The operator use case (incident response,
debugging) always wants the raw event log; there is no human-text rendering.
--format json and --format ndjson behave identically here; --format text is
silently treated as NDJSON.`,
Example: ` weknora session resume sess_xyz --message msg_abc
weknora session resume sess_xyz -m msg_abc --format ndjson`,
Args: cobra.ExactArgs(1),
RunE: func(c *cobra.Command, args []string) error {
opts.SessionID = args[0]
fopts, err := cmdutil.CheckFormatFlag(c)
if err != nil {
return err
}
fopts.ResolveDefault(iostreams.IO.IsStdoutTTY())
cli, err := f.Client()
if err != nil {
return err
}
return runResume(c.Context(), opts, fopts, cli)
},
}
cmd.Flags().StringVarP(&opts.MessageID, "message", "m", "",
"Assistant message ID to resume (from the init or agent_query event of the original stream)")
_ = cmd.MarkFlagRequired("message")
cmdutil.AddFormatFlag(cmd, resumeFields...)
cmdutil.SetAgentHelp(cmd, cmdutil.AgentHelp{
UsedFor: "Resume an SSE event stream for an in-progress or completed assistant message. Produces an NDJSON event stream: init line (session_id, message_id) then raw SDK StreamResponse events.",
RequiredFlags: []string{"<session-id> (positional)", "--message (persisted assistant message id — get it from `weknora message list --session <id>`; a live stream's assistant_message_id is not resumable once the message persists)"},
Examples: []string{
"weknora session resume sess_xyz --message msg_abc --format json",
"# Get the message id from: weknora message list --session <session-id> (the persisted assistant message)",
},
Output: "NDJSON stream: {type:init, session_id, message_id, profile} then SDK StreamResponse events (response_type, content, done, knowledge_references, assistant_message_id, ...)",
Warnings: []string{
"Server replays from event 0 (NOT cursor-from-disconnect). Agents that already consumed events on the original stream MUST dedupe by message_id + event hash to avoid double-processing.",
"Buffer TTL: redis mode 1h hardcoded; memory mode = process lifetime. After expiry the CLI returns local.sse_stream_aborted.",
"Output is always NDJSON (an event stream, not an envelope): --jq does not apply and --format text/json/ndjson behave identically here — parse the event lines yourself.",
},
})
return cmd
}
// runResume is the testable core: validate, dispatch the resume, and
// route the NDJSON stream. Returns a typed error.
//
// Always emits NDJSON: a buffered envelope makes no sense for a streaming
// command, and resume has no human-text use case (operators reach
// for it during incident response / debugging, which always wants the raw
// event log). --format text is therefore treated identically to --format
// json/ndjson here.
func runResume(ctx context.Context, opts *ResumeOptions, _ *cmdutil.FormatOptions, svc ResumeService) error {
if opts.SessionID == "" {
return cmdutil.NewError(cmdutil.CodeInputInvalidArgument, "session-id argument cannot be empty")
}
if opts.MessageID == "" {
return cmdutil.NewError(cmdutil.CodeInputInvalidArgument, "--message cannot be empty")
}
if svc == nil {
return cmdutil.NewError(cmdutil.CodeServerError, "session resume: no SDK client available")
}
w := iostreams.IO.Out
// 1. Inject the CLI-managed init event at stream head. Both session_id
// and message_id are populated so agents can key their dedupe table
// on the resumed message before the first SDK frame arrives.
initEv := output.InitEvent{
SessionID: opts.SessionID,
MessageID: opts.MessageID,
Profile: cmdutil.GetProfile(),
}
if err := output.EmitInit(w, initEv); err != nil {
return err
}
// 2. Open the SDK replay stream and pass each event through as a bare
// NDJSON line. The SDK's StreamResponse is the source of truth for
// the event vocabulary; the CLI does not reshape it.
// The SDK invokes the callback for each event (including a terminal
// response_type=error frame) BEFORE returning, so raw passthrough still
// emits every event; on a terminal error frame the SDK then returns an
// *SSEStreamError. No CLI-level early-terminate is needed.
cb := func(r *sdk.StreamResponse) error {
return output.EmitSDKEvent(w, r)
}
err := svc.ContinueStream(ctx, opts.SessionID, opts.MessageID, cb)
if err != nil {
// Ctrl-C / SIGTERM lineage (operator gave up on the resume).
if cmdutil.IsCancelled(ctx, err) {
return cmdutil.Wrapf(cmdutil.CodeOperationCancelled, err, "session resume cancelled")
}
// WrapStream routes through ClassifySDKError: a terminal SSE error
// frame classifies as server.error (matching chat / session ask); a
// pre-stream HTTP failure (e.g. 404 for an unknown message_id) still
// surfaces via ClassifyHTTPError as resource.not_found etc.
return cmdutil.WrapStream(err, "resume stream")
}
return nil
}
// compile-time check: production SDK client satisfies ResumeService.
var _ ResumeService = (*sdk.Client)(nil)