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.
This commit is contained in:
nullkey
2026-07-02 15:53:51 +08:00
committed by lyingbug
parent 040a734766
commit d3c03b2f34
95 changed files with 1711 additions and 436 deletions
+18 -13
View File
@@ -12,6 +12,7 @@ import (
"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"
)
@@ -26,17 +27,12 @@ var docCreateFields = []string{
// CreateOptions holds CLI flag values for `doc create`.
type CreateOptions struct {
Text string // --text (required): document text content (Markdown)
Title string // --title: document title (preferred; matches `doc update --title`)
Name string // --name: deprecated alias for --title
Title string // --title: document title (matches `doc update --title`)
TagID string // --tag-id: associate with a tag
Channel string // --channel: ingestion-channel tag (default "api")
DryRun bool
}
// title returns the resolved document title, preferring --title over the
// deprecated --name alias.
func (o *CreateOptions) title() string { return cmp.Or(o.Title, o.Name) }
// CreateService is the narrow SDK surface for `doc create`.
// *sdk.Client satisfies it.
type CreateService interface {
@@ -84,7 +80,7 @@ don't require a file upload or remote URL. KB resolution follows the standard
Action: "doc.create",
Args: map[string]any{
"text": opts.Text,
"title": opts.title(),
"title": opts.Title,
"kb": kbID,
},
}); handled {
@@ -106,8 +102,6 @@ don't require a file upload or remote URL. KB resolution follows the standard
cmdutil.AddKBFlag(cmd)
cmd.Flags().StringVar(&opts.Text, "text", "", "Document text content in Markdown format (required)")
cmd.Flags().StringVar(&opts.Title, "title", "", "Document title")
cmd.Flags().StringVar(&opts.Name, "name", "", "Document title (deprecated: use --title)")
_ = cmd.Flags().MarkDeprecated("name", "use --title instead")
cmd.Flags().StringVar(&opts.TagID, "tag-id", "", "Tag id to associate with the new entry")
cmd.Flags().StringVar(&opts.Channel, "channel", "", "Ingestion-channel tag recorded server-side (default \"api\")")
_ = cmd.MarkFlagRequired("text")
@@ -134,7 +128,7 @@ func runCreate(ctx context.Context, opts *CreateOptions, fopts *cmdutil.FormatOp
return cmdutil.NewFlagError(fmt.Errorf("--text is required"))
}
req := &sdk.CreateManualKnowledgeRequest{
Title: opts.title(),
Title: opts.Title,
Content: opts.Text,
TagID: opts.TagID,
Channel: cmp.Or(opts.Channel, uploadChannel),
@@ -143,10 +137,18 @@ func runCreate(ctx context.Context, opts *CreateOptions, fopts *cmdutil.FormatOp
if err != nil {
return cmdutil.WrapHTTP(err, "create document")
}
if fopts.WantsJSON() {
return fopts.Emit(iostreams.IO.Out, k, nil)
// Inline-created docs land in parse_status=draft and are NOT auto-queued
// for parsing (unlike `doc upload`), so they aren't searchable until
// reparsed. Surface the next step so an agent doesn't `doc wait` into a
// timeout or `search` into empty results.
var meta *output.Meta
if k.ParseStatus == "draft" {
meta = &output.Meta{Hint: "document created in parse_status=draft (not yet indexed) — run `weknora doc reparse " + k.ID + "` to parse & make it searchable"}
}
displayed := opts.title()
if fopts.WantsJSON() {
return fopts.Emit(iostreams.IO.Out, k, meta)
}
displayed := opts.Title
if displayed == "" {
displayed = k.Title
}
@@ -154,5 +156,8 @@ func runCreate(ctx context.Context, opts *CreateOptions, fopts *cmdutil.FormatOp
displayed = k.ID
}
fmt.Fprintf(iostreams.IO.Out, "✓ Created %q (id: %s)\n", displayed, k.ID)
if meta != nil {
fmt.Fprintf(iostreams.IO.Out, " ⚠ %s\n", meta.Hint)
}
return nil
}
+6 -13
View File
@@ -35,28 +35,21 @@ func (f *fakeCreateSvc) CreateManualKnowledge(
return f.resp, f.err
}
// TestCreate_TitlePreferredOverName: --title is the canonical flag; the
// deprecated --name remains a working alias but --title wins when both are set.
func TestCreate_TitlePreferredOverName(t *testing.T) {
// TestCreate_TitleSetsRequestTitle: --title is the sole title flag and flows
// straight into the create request.
func TestCreate_TitleSetsRequestTitle(t *testing.T) {
_, _ = iostreams.SetForTest(t)
svc := &fakeCreateSvc{resp: &sdk.Knowledge{ID: "d1"}}
require.NoError(t, runCreate(context.Background(),
&CreateOptions{Text: "x", Title: "FromTitle", Name: "FromName"},
&CreateOptions{Text: "x", Title: "FromTitle"},
&cmdutil.FormatOptions{Mode: cmdutil.FormatText}, svc, "kb1"))
assert.Equal(t, "FromTitle", svc.got.req.Title)
// --name alone still works (back-compat for the deprecated alias).
svc2 := &fakeCreateSvc{resp: &sdk.Knowledge{ID: "d2"}}
require.NoError(t, runCreate(context.Background(),
&CreateOptions{Text: "x", Name: "OnlyName"},
&cmdutil.FormatOptions{Mode: cmdutil.FormatText}, svc2, "kb1"))
assert.Equal(t, "OnlyName", svc2.got.req.Title)
}
func TestCreate_Success_Text(t *testing.T) {
out, _ := iostreams.SetForTest(t)
svc := &fakeCreateSvc{resp: &sdk.Knowledge{ID: "doc_manual_1", Title: "Sprint Notes"}}
opts := &CreateOptions{Text: "# Sprint Notes\n\nAction items: ...", Name: "Sprint Notes"}
opts := &CreateOptions{Text: "# Sprint Notes\n\nAction items: ...", Title: "Sprint Notes"}
require.NoError(t, runCreate(context.Background(), opts, &cmdutil.FormatOptions{Mode: cmdutil.FormatText}, svc, "kb_xxx"))
assert.Equal(t, "kb_xxx", svc.got.kbID)
@@ -110,7 +103,7 @@ func TestCreate_Channel_DefaultIsAPI(t *testing.T) {
func TestCreate_JSON_Envelope(t *testing.T) {
out, _ := iostreams.SetForTest(t)
svc := &fakeCreateSvc{resp: &sdk.Knowledge{ID: "doc_manual_json", Title: "My Note"}}
opts := &CreateOptions{Text: "# My Note", Name: "My Note"}
opts := &CreateOptions{Text: "# My Note", Title: "My Note"}
fopts := &cmdutil.FormatOptions{Mode: cmdutil.FormatJSON}
require.NoError(t, runCreate(context.Background(), opts, fopts, svc, "kb_xxx"))
+5
View File
@@ -79,11 +79,16 @@ stdout.`,
}
cmd.Flags().StringVarP(&opts.Output, "output", "O", "", `Output path; "-" for stdout. Defaults to the server-suggested filename.`)
cmd.Flags().BoolVar(&opts.Clobber, "clobber", false, "Overwrite the output file if it already exists")
cmdutil.AddIgnoredKBFlag(cmd)
cmdutil.AddFormatFlag(cmd, downloadFields...)
cmdutil.SetAgentHelp(cmd, cmdutil.AgentHelp{
UsedFor: "Download a document's bytes by id. Writes a file (or stdout with --output -).",
RequiredFlags: []string{"<doc-id> (positional)"},
Output: "with --format json (file output): envelope.data has path, bytes, filename; suppressed with --output - (raw bytes to stdout)",
Examples: []string{
"weknora doc download doc_abc --output ./manual.pdf",
"weknora doc download doc_abc --output - > manual.pdf",
},
})
return cmd
}
+2 -2
View File
@@ -106,7 +106,7 @@ backend storage order is not guaranteed and varies between deployments.`,
cmd.Flags().IntVarP(&opts.Limit, "limit", "L", 30, "Maximum results to return — client-side cap; meta.has_more/total_count report the full size (1..10000)")
cmd.Flags().BoolVar(&opts.AllPages, "all-pages", false, "Walk all server pages until exhausted (or --limit hit)")
cmd.Flags().StringVar(&opts.Status, "status", "", "Filter by parse status: pending | processing | completed | failed")
cmd.Flags().StringVar(&opts.Keyword, "keyword", "", "Server-side substring match against title / file_name (case-sensitive)")
cmd.Flags().StringVar(&opts.Keyword, "keyword", "", "Server-side substring match against title / file_name (case-insensitive)")
cmd.Flags().StringVar(&opts.FileType, "file-type", "", `Filter by file extension (e.g. "pdf", "md")`)
cmd.Flags().StringVar(&opts.Source, "source", "", `Filter by ingestion source (e.g. "api", "web")`)
cmd.Flags().StringVar(&opts.TagID, "tag-id", "", "Filter by tag association")
@@ -116,7 +116,7 @@ backend storage order is not guaranteed and varies between deployments.`,
cmdutil.SetAgentHelp(cmd, cmdutil.AgentHelp{
UsedFor: "List documents in the resolved knowledge base. Results come with meta.count; use --limit to cap, --all-pages to walk every server page, --status/--keyword to filter server-side.",
Examples: []string{"weknora doc list --format json", "weknora doc list --all-pages --limit 200 --format json"},
Output: "envelope.data is an array of Knowledge objects with id, title, file_name, parse_status; meta.count is the total returned; meta.total_count is the server-side total before client-side --limit truncation",
Output: "envelope.data is an array of Knowledge objects with id, title, file_name, parse_status; meta.count is the returned count; meta.total_count is the server-side total before client-side --limit truncation; meta.has_more=true when --limit truncated",
})
return cmd
}
+1
View File
@@ -57,6 +57,7 @@ keeps its id; parsing restarts asynchronously, so follow with
return runReparse(c.Context(), opts, fopts, cli, args[0])
},
}
cmdutil.AddIgnoredKBFlag(cmd)
cmdutil.AddFormatFlag(cmd, docReparseFields...)
cmdutil.AddDryRunFlag(cmd, &opts.DryRun)
cmdutil.SetAgentHelp(cmd, cmdutil.AgentHelp{
+1
View File
@@ -98,6 +98,7 @@ to the user first.`,
}
cmd.Flags().StringVar(&title, "title", "", "New title (omit to leave unchanged)")
cmd.Flags().StringVar(&desc, "description", "", "New description (omit to leave unchanged)")
cmdutil.AddIgnoredKBFlag(cmd)
cmdutil.AddFormatFlag(cmd, docUpdateFields...)
cmdutil.AddDryRunFlag(cmd, &opts.DryRun)
cmdutil.SetWriteRisk(cmd, "doc.update")
+9 -8
View File
@@ -81,16 +81,11 @@ func runUploadRecursive(ctx context.Context, opts *UploadOptions, fopts *cmdutil
// uploaded captures per-path server results for successful uploads.
// Populated by the RunBatch closure; read by the resultFn below.
uploaded := make(map[string]uploadedFile, len(matches))
var firstFailCode cmdutil.ErrorCode
channel := cmp.Or(opts.Channel, uploadChannel)
outcomes, runErr := cmdutil.RunBatch(ctx, matches, func(ctx context.Context, p string) error {
k, err := svc.CreateKnowledgeFromFile(ctx, kbID, p, meta, opts.EnableMultimodel, "", channel, nil)
if err != nil {
code := cmdutil.ClassifyHTTPError(err)
if firstFailCode == "" {
firstFailCode = code
}
// Per-file progress lines are human progress signal; suppress
// under --format json so they don't precede the JSON object on stdout.
if !fopts.WantsJSON() {
@@ -133,9 +128,15 @@ func runUploadRecursive(ctx context.Context, opts *UploadOptions, fopts *cmdutil
// carries per-file detail; without Silent the root error handler would
// print to stderr in addition. ExitCode still walks Code so the typed
// exit-code-by-class contract holds.
code := firstFailCode
if code == "" {
code = cmdutil.ClassifyContextErr(ctx.Err())
// Any per-file failure collapses to operation.failed (exit 1) — the
// per-file batch envelope carries each file's typed error, which is the
// authoritative signal (matches doc/session/chunk batch delete). A
// cancelled / timed-out batch keeps its context class (124 / 130) so a
// permanent partial failure (e.g. a duplicate) is never misreported as a
// retryable 5xx that an agent would loop on.
code := cmdutil.CodeOperationFailed
if ctxErr := ctx.Err(); ctxErr != nil {
code = cmdutil.ClassifyContextErr(ctxErr)
}
return &cmdutil.Error{
Code: code,
+7 -4
View File
@@ -109,9 +109,12 @@ func TestUploadRecursive_PartialFailure_Exits1(t *testing.T) {
var typed *cmdutil.Error
require.ErrorAs(t, err, &typed)
// CodeServerError preserves the 500 classification of the underlying
// SDK error - the recursive wrapper just aggregates.
assert.Equal(t, cmdutil.CodeServerError, typed.Code)
// Any per-file failure aggregates to operation.failed (exit 1), matching the
// batch-delete contract — a partial failure (even a permanent one like a
// duplicate, which classifies as server.error per-file) must not surface as
// a retryable exit 7 an agent would loop on. Per-file codes live in the
// batch envelope.
assert.Equal(t, cmdutil.CodeOperationFailed, typed.Code)
got := out.String()
assert.Contains(t, got, "OK") // ok.pdf still succeeded
@@ -267,5 +270,5 @@ func TestUploadRecursive_JSON_BatchEnvelope(t *testing.T) {
var typed *cmdutil.Error
require.ErrorAs(t, err, &typed)
assert.True(t, typed.Silent, "JSON-path partial failure must be Silent")
assert.Equal(t, cmdutil.CodeServerError, typed.Code)
assert.Equal(t, cmdutil.CodeOperationFailed, typed.Code)
}
+1
View File
@@ -52,6 +52,7 @@ func NewCmdView(f *cmdutil.Factory) *cobra.Command {
},
}
cmdutil.AddFormatFlag(cmd, docViewFields...)
cmdutil.AddIgnoredKBFlag(cmd)
cmdutil.SetAgentHelp(cmd, cmdutil.AgentHelp{
UsedFor: "fetch one document's metadata by id",
RequiredFlags: []string{"<doc-id> (positional)"},
+9
View File
@@ -94,6 +94,7 @@ For fail-fast semantics, use shell composition:
cmd.Flags().DurationVar(&opts.Timeout, "timeout", 10*time.Minute, "Max wait time before exiting 124")
cmd.Flags().DurationVar(&opts.Interval, "interval", 2*time.Second, "Initial poll interval; exponential backoff capped at 15s + jitter")
cmdutil.AddFormatFlag(cmd)
cmdutil.AddIgnoredKBFlag(cmd)
cmdutil.SetAgentHelp(cmd, cmdutil.AgentHelp{
UsedFor: "block until one or more documents reach a terminal parse state (completed or failed), or --timeout elapses",
RequiredFlags: []string{"<doc-id>... (one or more positionals)"},
@@ -234,6 +235,14 @@ func waitForDocs(ctx context.Context, ids []string, svc WaitService, opts WaitOp
case "failed":
addFailed(FailedDoc{ID: id, Message: doc.ErrorMessage})
return
case "draft":
// draft = created but NOT queued for parsing (inline
// `doc create` leaves docs here; file `doc upload`
// auto-enqueues). It never progresses on its own, so
// waiting would hang to the --timeout (124). Fail fast with
// the exact unblock command instead of silently polling.
addFailed(FailedDoc{ID: id, Message: "parse_status=draft: not queued for parsing — run `weknora doc reparse " + id + "` to index it"})
return
}
// Not yet terminal — sleep with jitter, then exp-backoff.
+27
View File
@@ -311,3 +311,30 @@ func TestDocWait_FailureError_IsSilent(t *testing.T) {
t.Errorf("exit code = %d, want 1", cmdutil.ExitCode(err))
}
}
// TestWaitForDocs_DraftFailsFast pins that a document stuck in parse_status
// "draft" (inline `doc create` leaves docs here; it never auto-progresses) is
// failed fast with a reparse hint instead of polling until the --timeout.
// Regression: `doc create` + `doc wait` used to hang to a 124 timeout.
func TestWaitForDocs_DraftFailsFast(t *testing.T) {
svc := newFakeKBSvc(map[string][]string{
"doc_draft": {"draft", "draft", "draft"},
})
start := time.Now()
res, _ := waitForDocs(context.Background(), []string{"doc_draft"}, svc, WaitOptions{
Timeout: 5 * time.Second,
Interval: 1 * time.Millisecond,
})
if len(res.Timeout) != 0 {
t.Errorf("draft must NOT time out; got timeout=%v", res.Timeout)
}
if len(res.Failed) != 1 || res.Failed[0].ID != "doc_draft" {
t.Fatalf("draft doc must be failed-fast; got failed=%v", res.Failed)
}
if !strings.Contains(res.Failed[0].Message, "reparse") {
t.Errorf("draft failure message must point to reparse; got %q", res.Failed[0].Message)
}
if elapsed := time.Since(start); elapsed > time.Second {
t.Errorf("draft must fail fast (well under timeout); took %v", elapsed)
}
}