Files
archie/internal/channels/telegram/richblocks.go
T
sam c4ddb9acdf Feat/q9au inline review comments (#895)
* feat(workflow): post line-anchored review comments with suggested fixes

Bead archie-core-q9au (docs/prds/inline-review.md). A review's findings reached
the operator only as a flat list in the PR body, so the author had to map every
finding back to its line by hand and apply mechanical fixes by hand. Now the
line-anchored findings are also posted as inline review comments, carrying a
one-click suggestion when the reviewer could state the fix.

Domain (internal/domain/workflow):
- ReviewFinding gains Suggestion; Validate rejects one on anything but a
  confirmed finding, so the agent's record_finding call is refused with
  something it can act on. Validate splits into validateAnchor /
  validateClassification / validateDisposition, which also brings it back under
  the cyclop limit the gate enforces.
- StagePostReviewComments runs after StageOpenPR in the implement workflow. It
  posts only line-anchored findings (Line > 0); a whole-file finding has no line
  to attach to and stays in the body list, which remains the complete record.
  It is best-effort: the PR is open and the body already lists every finding, so
  a failed comment is logged, never a park -- the same reasoning as OpenPR's
  existing best-effort LinkBranch call.
- The suggestion fence is gated on Verdict == confirmed where the fence is
  written, because it is a one-click apply: a wrong suggestion breaks the code
  and the author has to notice.

Forge (internal/forge): ReviewCommentWriter + GitHub/Gitea implementations,
type-asserted like PullRequestReader so a forge that cannot is skipped rather
than faked (NoopForge is pinned as not satisfying it). GitHub loops one
CreateComment per comment and joins the failures, so one refused anchor cannot
drop the rest; Gitea carries them on a single COMMENT-state review.

Two deliberate departures from the PRD's sketch, both recorded in it:
- The interface is batch (CreateReviewComments), not one call per comment:
  Gitea's only inline shape is one submitted review, and the set then crosses
  NATS once instead of once per finding.
- The head SHA is resolved inside each forge implementation rather than by the
  caller. The stage runs in archie-agent, which holds no forge credentials and
  has no PullRequestReader; the daemon-side forgerpc handler is the only place
  that does.

Transport (internal/forgerpc): workflow.Forger gains the method, so the agent
reaches the daemon over the new create_review_comments subject; the server
type-asserts the writer and answers an incapable forge with an error the stage
logs. This is the one proxied method whose absence on an image skew does not
park a task.

The reviewer prompt and record_finding schema ask for a suggestion on a
confirmed, mechanically fixable defect only.

Regenerating the wfextract symbol table -- the documented way a new exported
workflow symbol reaches interpreted stage plugins -- also picked up symbols that
were missing from the committed table (the Status* and Source* constants,
ReviewCheck, Store, Task, and the _Store wrapper). Nil-guards were re-applied to
every wrapper afterwards, as that file's header requires.

Evidence: task check exit 0 (incl. 229 UI tests). Three mutations of the new
stage logic each fail a test that names the behaviour -- dropping the
confirmed-only gate, dropping the Line > 0 filter, and letting a posting failure
escape -- so the tests discriminate rather than passing vacuously.

* fix(build): drop the Dockerfile pin assertion

The guard fetched golangci-lint's go.mod at image build time to fail the build
when GOLANGCI_LINT_VERSION and GOFUMPT_VERSION disagreed. It only ever tripped
on self-inflicted pin drift, but it added a network fetch and a new failure
mode to the release build, which is not where that belongs.

The pins stay correct (golangci-lint 2.13.2, whose vendored gofumpt v0.11.0
equals GOFUMPT_VERSION) and the comment above them still states the invariant,
so the reasoning survives without the build-time machinery. The pin change
itself reaches the agent image through the normal release path: CI rebuilds
and pushes archie-agent on an archie/vX.Y.Z release, and a gateway-only
release deliberately does not.

* feat(workflow): add the remediate workflow for PR review remediation

Implements step 5 of docs/prds/pr-review-remediation.md: a remediate
workflow that resumes the task's existing worktree onto its PR branch
(worktree.Manager.Resume, not Prepare's refresh), runs a builder agent
against the review's actionable comments, commits and pushes to the
same branch, and replies via the forge's ReplyToReview. Task.RetryCount
doubles as the remediation round counter, capped by the repo's
max_retries; the round-cap stage parks the task and posts one comment
when it's hit.

Task gains a ReviewPayload field (JSON-encoded review unit) with full
store/proto/staterpc plumbing. workflow.Forger gains Comment and
ReplyToReview, proxied to archie-agent via a new forgerpc RPC. Resume
and its Dir lookup are added as an unexported optional Trees capability
rather than widening the exported interface, which is already at
golangci-lint's interfacebloat cap and is projected into wfextract's
Yaegi symbol table for repository-authored stage plugins.

The daemon-side reaction consumer that dispatches into this workflow
(subscribing ARCHIE_REACTIONS, grouping comments by review, enforcing
per-task serialization) is tracked separately in archie-core-8li9.4.

* feat(messaging): extract standalone messaging service binary and process composition

* feat(messaging): adapt channel frontends to consume Gateway ChatContract

* feat(messaging): remove in-process channels and Router from daemon bootstrap

Telegram, email and webhook now run only in the Messaging Service, which
reaches this process through the Gateway's ChatContract. The daemon builds
no channel and no gateway.Router, so the wiring that existed only to serve
them goes with it: setupGateways, setupEmailGateway, setupWebhookGateway,
setupTelegramGateway, buildTelegramRouter, makeTelegramReload and the
release-announcement path helper.

The seams those paths owned move to where they can still be honoured:

- Channel ValidateConfig runs in archiemessaging.compose, which now returns
  an error, so a configured-but-invalid front-end fails the service rather
  than being dropped silently.
- The inbound rate limiter moves to the Gateway process, which owns the only
  Router and therefore the only place a per-sender budget applies to every
  channel's turns.
- The daemon's gateway readiness probe becomes a ContractProbe against the
  Gateway. It cannot observe a channel lifecycle it no longer runs, and a
  channel-count probe would have reported every daemon degraded.
- telegramSetup is now chatSetup (in chat_setup.go), trimmed to the fields a
  chat turn runner actually reads.

Telegram operator features that were wired only in the daemon (/update,
release announcements, version reporting, SIGHUP reload) are not yet wired
in the Messaging Service: archie-core-8cda.6.6. Channel lifecycle reporting:
archie-core-8cda.6.8. The Gateway process builds no worktree manager, so
review_pr is unadvertised on every channel: archie-core-8cda.6.7.

* feat(messaging): re-wire the Telegram operator surface in the Messaging Service

setupTelegramGateway satisfied eight telegram.Gateway seams from the daemon's
own process. Each is resolved by where the fact it needs is owned, not by
where it used to be built; the decision table is in
docs/architecture/migration-decisions.md.

Wired here: /update (built from this service's own [chat.telegram] check and
install commands), the update-report relay (same sha256(bot_user)[:8] path the
daemon wrote, so a pending report is found rather than written afresh),
show_tool_calls, and /restart's config reload, which re-resolves the token and
allowlist from this service's own file and refuses a reload that resolves to
an empty token.

/version is sourced from the Gateway: ChatSnapshot gains a Version field
carrying the Gateway's rendered build block, read per call and bounded by the
dependency timeout. cmd/archie-messaging deliberately takes no version
ldflags, so it cannot report a partial upgrade as a matched one.

RunningVersions and ReleaseAnnouncements stay nil. Both need a component's
real running version, and the only versions this process can obtain are a
different binary's stamps -- reporting them would manufacture exactly the
false success the update check exists to catch, and the announcer without
parseable versions is a no-op that reads as configured. archie-core-8cda.6.9
carries the contract that unblocks them. Dangerous stays nil: no composition
has ever supplied a command authority.

Enrich is dropped from the update service: componentInstallTypeEnricher read
[nats], which this service must not decode, so the check command's own
reported install type stands.

* fix(workflow): guard inline review comments against head drift

Two findings from the review of q9au.

A comment's line numbers describe exactly one revision of the branch, but the
forge resolved the pull request's head at posting time -- after the review had
already read the worktree. A push in that window (a collaborator, a retry) meant
the findings were anchored to whatever now occupied those lines, mislabelling
instead of failing. The revision the review measured now travels with the
comment set: OpenPR records it from the same worktree read it already performs
(TaskContext.ReviewedHeadSHA), StagePostReviewComments threads it through
Forger to forgerpc, and the forge refuses the whole set once the head it reads
has moved off it -- which the best-effort stage logs, the PR body remaining the
record. An unmeasured revision posts unverified rather than dropping every
finding, so a worktree that cannot report one degrades instead of regressing.

Suggestion was documented single-line only, because GitHub applies a suggestion
fence up to its first line and a multi-line replacement would silently corrupt
the file. Only the verdict was validated, so a multi-line suggestion reached the
renderer and posted as that truncated fence. Validate now rejects an embedded
newline, which makes record_finding refuse it with feedback the model can act on
instead of the reviewer discovering the corruption later.

* fix(gateway): render chat reports as lines, not one run-on paragraph

/status and /tasks are line-oriented reports, but their newlines were plain
"\n" -- CommonMark's soft wrap. The Telegram rich-block renderer space-joins
soft-wrapped lines, by design: a Telegram paragraph renders an embedded
newline as nothing at all, so a space is the only separator that survives.
The reports therefore arrived as one line each: "Runtime Provider: DeepSeek
Model: deepseek/deepseek-flash", and every task's title, state and park
reason jammed together.

The renderer now implements CommonMark's hard line break (two trailing
spaces, or a trailing backslash): the line closes its own block and the next
one hugs it, so a stanza keeps its lines without the spacer paragraph that
separates unrelated blocks. Ordinary prose is untouched -- a model's wrapped
paragraph still joins with a space. The producers mark their report lines
with it.

Two report fixes alongside:

- A park reason is capped to its first line at 160 runes. It carries whatever
  the failing stage reported, which for a lint gate is the whole tool log;
  unbounded, one parked task pushed every other task out of the reply. The
  full reason is still on the task.
- /status loses its "Runtime" sub-header. A section header over two fields
  made a four-fact report read like a document.
2026-09-19 23:44:51 +10:00

583 lines
19 KiB
Go

package telegram
import (
"regexp"
"strings"
"unicode"
"unicode/utf8"
"github.com/go-telegram/bot/models"
)
// Telegram's rich-message renderer has no ATX heading support and its
// MarkdownV2 dialect needs escaping, so rather than translating Markdown into a
// format it only partially implements we emit structured blocks the renderer
// supports natively. Inline emphasis/code/links are reduced to plain text so no
// Markdown marker can survive into the rendered message.
var (
inlineLinkHTML = regexp.MustCompile(`\[([^\]]*)\]\(([^)]+)\)`)
codeBlockMark = regexp.MustCompile("`([^`]*)`")
)
// emphasisMarkers are the delimiters stripEmphasis pairs, longest first so
// "**" is consumed before the "*" pass can see it as two single markers.
var emphasisMarkers = []string{"~~", "**", "__", "*", "_"}
// richText builds a plain Text; an empty RichText.Type serializes as a JSON
// string, which is exactly what a plain-text rich block wants.
func richText(text string) models.RichText {
return models.RichText{PlainText: text}
}
func paragraphBlock(text string) models.InputRichBlock {
return models.InputRichBlock{
Type: models.RichBlockTypeParagraph,
InputRichBlockParagraph: &models.InputRichBlockParagraph{
Type: models.RichBlockTypeParagraph,
Text: richText(text),
},
}
}
func headingBlock(line string) models.InputRichBlock {
depth := headingDepth(line)
text := stripInlineMarkdown(strings.TrimSpace(line[depth:]))
return models.InputRichBlock{
Type: models.RichBlockTypeSectionHeading,
InputRichBlockSectionHeading: &models.InputRichBlockSectionHeading{
Type: models.RichBlockTypeSectionHeading,
Text: richText(text),
Size: depth,
},
}
}
func preformattedBlock(text, lang string) models.InputRichBlock {
return models.InputRichBlock{
Type: models.RichBlockTypePreformatted,
InputRichBlockPreformatted: &models.InputRichBlockPreformatted{
Type: models.RichBlockTypePreformatted,
Text: richText(text),
Language: lang,
},
}
}
func headingDepth(line string) int {
n := 0
for n < len(line) && n < 6 && line[n] == '#' {
n++
}
if n == 0 {
return 0
}
if n < len(line) && line[n] != ' ' {
return 0
}
return n
}
var (
bulletListMarkers = []string{"- ", "* ", "+ "}
orderedListMark = regexp.MustCompile(`^[0-9]+[.)] `)
)
func isListItem(line string) bool {
for _, marker := range bulletListMarkers {
if strings.HasPrefix(line, marker) {
return true
}
}
return orderedListMark.MatchString(line)
}
func listItemText(line string) string {
for _, marker := range bulletListMarkers {
if strings.HasPrefix(line, marker) {
return stripInlineMarkdown(strings.TrimSpace(line[len(marker):]))
}
}
if loc := orderedListMark.FindStringIndex(line); loc != nil {
return stripInlineMarkdown(strings.TrimSpace(line[loc[1]:]))
}
return ""
}
// stripInlineMarkdown reduces inline emphasis, inline code and links to plain
// text, dropping the markers so none of them can leak into the rendered block.
func stripInlineMarkdown(s string) string {
s = inlineLinkHTML.ReplaceAllString(s, "$1 ($2)")
for _, marker := range emphasisMarkers {
s = stripEmphasis(s, marker)
}
return codeBlockMark.ReplaceAllString(s, "$1")
}
// stripEmphasis removes paired emphasis delimiters and keeps their content.
// Pairing follows CommonMark's flanking rules, so a delimiter that cannot
// open or close is ordinary text and stays put. That is what keeps this
// project's own vocabulary intact over the wire: snake_case config keys,
// dotted paths like config_schema_test.go, and Go pointer types such as
// *sql.Tx all used to lose characters to a non-greedy `_(.+?)_` that paired
// any two markers on a line (archie-core-8x1r).
func stripEmphasis(s, marker string) string {
var b strings.Builder
for i := 0; i < len(s); {
if !strings.HasPrefix(s[i:], marker) {
b.WriteByte(s[i])
i++
continue
}
end := i + len(marker)
if canOpen, _ := flanking(s, i, end, marker); canOpen {
if closer := findCloser(s, end, marker); closer >= 0 {
b.WriteString(s[end:closer])
i = closer + len(marker)
continue
}
}
b.WriteString(marker)
i = end
}
return b.String()
}
// findCloser returns the offset of the first delimiter at or after from that
// may close emphasis, or -1 when the run is unpaired.
func findCloser(s string, from int, marker string) int {
for i := from; i+len(marker) <= len(s); i++ {
if !strings.HasPrefix(s[i:], marker) {
continue
}
if _, canClose := flanking(s, i, i+len(marker), marker); canClose {
return i
}
}
return -1
}
// delimiterFlanks carries CommonMark's flanking facts about one delimiter
// run: whether it is left- and/or right-flanking, and whether the characters
// either side are punctuation, which the underscore rules also need.
type delimiterFlanks struct {
left, right bool
prevPunct, nextPunct bool
}
// flanksOf computes the flanking facts for the delimiter run s[i:j].
func flanksOf(s string, i, j int) delimiterFlanks {
prev := ' '
if i > 0 {
prev, _ = utf8.DecodeLastRuneInString(s[:i])
}
next := ' '
if j < len(s) {
next, _ = utf8.DecodeRuneInString(s[j:])
}
prevSpace, nextSpace := unicode.IsSpace(prev), unicode.IsSpace(next)
f := delimiterFlanks{prevPunct: isPunct(prev), nextPunct: isPunct(next)}
f.left = !nextSpace && (!f.nextPunct || prevSpace || f.prevPunct)
f.right = !prevSpace && (!f.prevPunct || nextSpace || f.nextPunct)
return f
}
// flanking reports whether the delimiter run s[i:j] may open or close
// emphasis. The rules are CommonMark's, including the one that an underscore
// may do neither inside a word.
//
// Single "*" carries one deliberate deviation: it may not open on
// punctuation. CommonMark reads "*.go files in cmd/*" as emphasis, which in a
// relay carrying shell globs and pointer types corrupts far more text than a
// stray marker would. Emphasis opening on a word still works.
func flanking(s string, i, j int, marker string) (canOpen, canClose bool) {
f := flanksOf(s, i, j)
switch marker {
case "_", "__":
return f.left && (!f.right || f.prevPunct), f.right && (!f.left || f.nextPunct)
case "*":
return f.left && !f.nextPunct, f.right
default:
return f.left, f.right
}
}
func isPunct(r rune) bool {
return unicode.IsPunct(r) || unicode.IsSymbol(r)
}
// markdownToBlocks converts a Markdown document into Telegram rich-message
// blocks. It is intentionally conservative: unknown constructs fall back to a
// plain paragraph rather than risking literal markers in the rendered output.
func markdownToBlocks(md string) []models.InputRichBlock {
if strings.TrimSpace(md) == "" {
return nil
}
p := &markdownBlockParser{}
for raw := range strings.SplitSeq(md, "\n") {
p.handleLine(strings.TrimRight(raw, "\r"))
}
p.closeCode()
p.flush()
return p.blocks
}
// markdownBlockParser holds the mutable state markdownToBlocks threads
// through its line-by-line scan: the accumulated blocks, and whichever
// multi-line construct (paragraph, code fence, list, block quote) is
// currently open.
type markdownBlockParser struct {
blocks []models.InputRichBlock
paragraph strings.Builder
codeLines []string
codeLang string
inCode bool
indented bool
listItems [][]models.InputRichBlock
quoteLines []string
// hardBreak records that the line just consumed ended in CommonMark's
// line-break marker, so the next line starts its own block.
hardBreak bool
// tightNext suppresses the spacer before the next block, keeping the
// lines of one hard-broken stanza visually together.
tightNext bool
}
// handleLine processes one line of input, dispatching to whichever
// construct is open or starting a new one.
func (p *markdownBlockParser) handleLine(line string) {
trimmed := strings.TrimSpace(line)
if p.inCode && p.consumeCodeLine(line, trimmed) {
return
}
switch {
case strings.HasPrefix(trimmed, "```"):
p.flush()
p.codeLang = strings.TrimSpace(strings.TrimPrefix(trimmed, "```"))
p.inCode = true
p.codeLines = nil
case headingDepth(trimmed) > 0:
p.flush()
p.appendBlock(headingBlock(trimmed))
case isListItem(trimmed):
p.flushParagraph()
p.flushQuote()
if len(p.listItems) == 0 {
p.flushList()
}
p.listItems = append(p.listItems, []models.InputRichBlock{paragraphBlock(listItemText(trimmed))})
case p.startsIndentedCode(line):
// CommonMark's other code form. Models emit it constantly, and
// without this it fell to the paragraph case below, which
// space-joins its lines into one run-on line (archie-core-cvu6).
p.flush()
p.inCode, p.indented, p.codeLang = true, true, ""
p.codeLines = []string{stripCodeIndent(line)}
case len(trimmed) > 1 && trimmed[0] == '>':
p.flushParagraph()
p.flushList()
p.quoteLines = append(p.quoteLines, line)
case trimmed == "":
p.flush()
default:
// A normal line ends a list or quote block and continues a
// paragraph. Joined with a space, not "\n": this is a CommonMark
// soft line break, and Telegram's rich-block paragraph text does
// not render an embedded "\n" as a break at all -- it renders
// nothing, jamming the two lines together with no separator
// whatsoever. A space is the correct soft-break rendering anyway
// and survives regardless of how the client treats a literal "\n".
//
// A line that ended in a hard break is the exception: the producer
// said it is a line, so it closes its own block and the next one
// hugs it. Line-oriented reports (/status, /tasks) depend on this;
// without it every one of their lines soft-joined into one run-on
// paragraph (archie-core-8cda.6.10).
p.flushList()
p.flushQuote()
switch {
case p.hardBreak:
p.flushParagraph()
p.tightNext = true
case p.paragraph.Len() > 0:
p.paragraph.WriteString(" ")
}
p.hardBreak = endsWithHardBreak(line)
p.paragraph.WriteString(trimHardBreak(trimmed))
}
}
// endsWithHardBreak reports whether line carries CommonMark's line-break
// marker: two or more trailing spaces, or a trailing backslash.
func endsWithHardBreak(line string) bool {
if strings.TrimSpace(line) == "" {
return false
}
if strings.HasSuffix(strings.TrimRight(line, " \t"), "\\") {
return true
}
return strings.HasSuffix(line, " ")
}
// trimHardBreak removes a trailing backslash marker, which unlike trailing
// spaces would otherwise survive into the rendered text.
func trimHardBreak(trimmed string) string {
return strings.TrimRight(strings.TrimSuffix(trimmed, "\\"), " ")
}
// consumeCodeLine handles one line while a code block is open, reporting
// whether that block claimed it. A fenced block ends at its closing fence; an
// indented block ends at the first non-indented, non-blank line, which is not
// a terminator and must still be parsed normally.
func (p *markdownBlockParser) consumeCodeLine(line, trimmed string) bool {
if !p.indented {
if strings.HasPrefix(trimmed, "```") {
p.closeCode()
} else {
p.codeLines = append(p.codeLines, line)
}
return true
}
if trimmed == "" || indentedCodeLine(line) {
p.codeLines = append(p.codeLines, stripCodeIndent(line))
return true
}
p.closeCode()
return false
}
// startsIndentedCode reports whether line opens an indented code block. It may
// not interrupt an open paragraph, list or quote, so a wrapped prose line that
// happens to be indented stays prose.
func (p *markdownBlockParser) startsIndentedCode(line string) bool {
return indentedCodeLine(line) && p.paragraph.Len() == 0 &&
len(p.listItems) == 0 && len(p.quoteLines) == 0
}
// flush closes every open construct, in the order that keeps blocks in
// source order.
func (p *markdownBlockParser) flush() {
p.flushList()
p.flushQuote()
p.flushParagraph()
}
// appendBlock adds block to the output. Telegram's rich-block renderer adds
// no vertical gap between adjacent blocks on its own -- only a heading
// carries its own margin -- so two blocks from source lines with no blank
// line between them (an intro line immediately followed by a list, two
// back-to-back paragraphs) render glued together with no visible break at
// all. An empty paragraph spacer before any non-heading-following block
// makes that gap explicit instead of relying on client-side spacing that
// doesn't exist.
func (p *markdownBlockParser) appendBlock(block models.InputRichBlock) {
tight := p.tightNext
p.tightNext = false
if !tight && len(p.blocks) > 0 && p.blocks[len(p.blocks)-1].Type != models.RichBlockTypeSectionHeading {
p.blocks = append(p.blocks, paragraphBlock(""))
}
p.blocks = append(p.blocks, block)
}
func (p *markdownBlockParser) flushParagraph() {
// A pending hard break belongs to the paragraph being closed. Left set, it
// would make the next block hug whatever came before a blank line.
p.hardBreak = false
if p.paragraph.Len() == 0 {
return
}
if text := strings.TrimSpace(p.paragraph.String()); text != "" {
p.appendBlock(paragraphBlock(stripInlineMarkdown(text)))
}
p.paragraph.Reset()
}
func (p *markdownBlockParser) flushList() {
if len(p.listItems) == 0 {
return
}
items := make([]models.InputRichBlockListItem, 0, len(p.listItems))
for _, item := range p.listItems {
items = append(items, models.InputRichBlockListItem{Blocks: item})
}
p.appendBlock(models.InputRichBlock{
Type: models.RichBlockTypeList,
InputRichBlockList: &models.InputRichBlockList{
Type: models.RichBlockTypeList,
Items: items,
},
})
p.listItems = nil
}
func (p *markdownBlockParser) flushQuote() {
if len(p.quoteLines) == 0 {
return
}
var inner []models.InputRichBlock
for _, line := range p.quoteLines {
body := strings.TrimSpace(strings.TrimPrefix(strings.TrimSpace(line), ">"))
if body == "" {
continue
}
inner = append(inner, paragraphBlock(stripInlineMarkdown(body)))
}
if len(inner) > 0 {
p.appendBlock(models.InputRichBlock{
Type: models.RichBlockTypeBlockQuotation,
InputRichBlockBlockQuotation: &models.InputRichBlockBlockQuotation{
Type: models.RichBlockTypeBlockQuotation,
Blocks: inner,
},
})
}
p.quoteLines = nil
}
// closeCode emits the open code block. The text is passed through verbatim:
// code is not prose, and running the emphasis stripper over it turned
// *sql.Tx into sql.Tx and ate same-line snake_case pairs (archie-core-8x1r).
func (p *markdownBlockParser) closeCode() {
lines := p.codeLines
if p.indented {
for len(lines) > 0 && strings.TrimSpace(lines[len(lines)-1]) == "" {
lines = lines[:len(lines)-1]
}
}
if len(lines) > 0 || p.codeLang != "" {
p.appendBlock(preformattedBlock(strings.Join(lines, "\n"), p.codeLang))
}
p.codeLines, p.codeLang, p.inCode, p.indented = nil, "", false, false
}
// indentedCodeLine reports whether line opens or continues an indented code
// block: four spaces or a tab, and something other than whitespace after.
func indentedCodeLine(line string) bool {
if strings.TrimSpace(line) == "" {
return false
}
return strings.HasPrefix(line, " ") || strings.HasPrefix(line, "\t")
}
// stripCodeIndent removes the one level of indentation that marks the block,
// so the code reads at its own natural margin.
func stripCodeIndent(line string) string {
if strings.HasPrefix(line, "\t") {
return line[1:]
}
return strings.TrimPrefix(line, " ")
}
// blocksToPlainText flattens blocks into readable plain text for the fallback
// send, joining each block's text with a blank line. Because it derives from
// the structured blocks there are no Markdown markers left to strip.
func blocksToPlainText(blocks []models.InputRichBlock) []string {
var out []string
for _, block := range blocks {
switch block.Type {
case models.RichBlockTypeParagraph, models.RichBlockTypeSectionHeading, models.RichBlockTypePreformatted,
models.RichBlockTypeFooter, models.RichBlockTypePullQuotation:
if text := blockPlainText(block); text != "" {
out = append(out, text)
}
case models.RichBlockTypeList:
if text := listBlockPlainText(block); text != "" {
out = append(out, text)
}
case models.RichBlockTypeBlockQuotation:
if text := quoteBlockPlainText(block); text != "" {
out = append(out, text)
}
case models.RichBlockTypeDivider:
out = append(out, "———")
}
}
return out
}
// listBlockPlainText renders a list block's items as newline-joined plain
// text, or "" if it has none.
func listBlockPlainText(block models.InputRichBlock) string {
if block.InputRichBlockList == nil {
return ""
}
var items []string
for _, item := range block.InputRichBlockList.Items {
for _, inner := range item.Blocks {
if text := blockPlainText(inner); text != "" {
items = append(items, text)
}
}
}
return strings.Join(items, "\n")
}
// quoteBlockPlainText renders a block quotation's inner blocks as
// newline-joined plain text, or "" if it has none.
func quoteBlockPlainText(block models.InputRichBlock) string {
if block.InputRichBlockBlockQuotation == nil {
return ""
}
var inner []string
for _, b := range block.InputRichBlockBlockQuotation.Blocks {
if text := blockPlainText(b); text != "" {
inner = append(inner, text)
}
}
return strings.Join(inner, "\n")
}
func blockPlainText(block models.InputRichBlock) string {
switch block.Type {
case models.RichBlockTypeParagraph:
if block.InputRichBlockParagraph != nil {
return block.InputRichBlockParagraph.Text.PlainText
}
case models.RichBlockTypeSectionHeading:
if block.InputRichBlockSectionHeading != nil {
return block.InputRichBlockSectionHeading.Text.PlainText
}
case models.RichBlockTypePreformatted:
if block.InputRichBlockPreformatted != nil {
return block.InputRichBlockPreformatted.Text.PlainText
}
case models.RichBlockTypeFooter:
if block.InputRichBlockFooter != nil {
return block.InputRichBlockFooter.Text.PlainText
}
case models.RichBlockTypePullQuotation:
if block.InputRichBlockPullQuotation != nil {
return block.InputRichBlockPullQuotation.Text.PlainText
}
}
return ""
}
// splitBlocks divides blocks into chunks that each fit within maxLen runes,
// ending a chunk only at a block boundary so a code block or heading is never
// cut in half. Every block is retained; ordering is preserved. A single block
// larger than maxLen is returned on its own so it is not silently dropped.
func splitBlocks(blocks []models.InputRichBlock, maxLen int) [][]models.InputRichBlock {
if len(blocks) == 0 || maxLen <= 0 {
return [][]models.InputRichBlock{blocks}
}
var chunks [][]models.InputRichBlock
var current []models.InputRichBlock
var currentLen int
for _, block := range blocks {
blockLen := len([]rune(strings.Join(blocksToPlainText([]models.InputRichBlock{block}), "\n")))
if len(current) > 0 && currentLen+blockLen > maxLen {
chunks = append(chunks, current)
current, currentLen = nil, 0
}
current = append(current, block)
currentLen += blockLen
}
if len(current) > 0 {
chunks = append(chunks, current)
}
return chunks
}