Perplexity's API stopped including Access-Control-Allow-Origin headers
for app://obsidian.md, so Electron's Chromium renderer began blocking
every streaming response body — even on successful 200 OK requests.
Every Perplexity query that used streaming was silently producing nothing.
Both streaming call sites now use https.request from node:https instead
of activeWindow.fetch. Node.js routes through the OS network stack and
has no CORS constraint at all. The resulting IncomingMessage stream is
wrapped in a Web ReadableStream<Uint8Array> and passed to the existing
SSE parsing machinery unchanged — idle-timeout, chunk accumulation,
citations, images, and AbortController signal handling all preserved.
Non-streaming requests (Obsidian's request() function) were never
affected and are untouched.
perplexityService.ts:
- Added import * as https from 'node:https' + import type { IncomingMessage }
- Replaced activeWindow.fetch in queryPerplexity streaming branch with
https.request → IncomingMessage → ReadableStream<Uint8Array> bridge
directoryTemplateService.ts:
- Same imports
- Replaced activeWindow.fetch in streamPerplexityToFile with same bridge
- AbortController.signal threaded into https.request signal option so
ceiling-timer and user-cancel abort behavior is fully preserved
Bumps to 0.3.1 — manifest.json, package.json, versions.json updated.
Files changed:
- src/services/perplexityService.ts
- src/services/directoryTemplateService.ts
- manifest.json
- package.json
- versions.json
- changelog/2026-07-06_01.md
- changelog/releases/0.3.1.md
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
0.3.0 turns Perplexed into a serious analyst tooling layer for Venture
Capital, Private Equity, and equities-trading workflows. Three new directory
templates produce 6-9K-word cited analyst drafts in a single Perplexity
Deep Research run. Underneath them, four infrastructure fixes turn
deep-research from a brittle "might work" workflow into something you can
put real analyst time behind.
The three new templates: market-map-profile (analyst memo for Known
Category or Thesis-Driven maps), standards-and-specs-profile (open-spec
profiles with five-way authority typing and three-tier adoption framing),
market-category-profile (concept-folder reference card with three
financial-stage tiers — Incumbents / Challengers / Innovators). The four
infrastructure fixes: per-template request-timeout-ms wall-clock override
(yesterday, pressure-relief valve), structural port of per-chunk
idle-timeout discipline from the legacy modal flow into the
directory-template flow (yesterday, structural fix), per-template
max-tokens Perplexity output-budget override (today, fixes silent
clean-truncation when finish_reason=length fires mid-template), and a
strengthened mermaid-discipline partial plus new latex-discipline partial
(today, fixes the 4-of-5 broken-Mermaid rate via paired BAD/GOOD examples
and a six-item pre-emit self-check).
New templates (src/docs/templates/):
- market-map-profile.md — sonar-deep-research, request-timeout-ms 2400000
(40-min ceiling), max-tokens 24000. Anti-incumbent editorial stance.
- standards-and-specs-profile.md — sonar-deep-research, request-timeout-ms 0
(idle-only safety), max-tokens 24000. Five-way authority typing, named
editors, stewardship transitions, named critics.
- market-category-profile.md — sonar-deep-research, request-timeout-ms 0,
max-tokens 24000. Three-tier financial-stage framing, separate Why Now
and What's Happening sections, Industry Coverage sub-grouped into
Market Reports / Industry Articles / Financial News.
Rendering-discipline partials (src/docs/partials/):
- mermaid-discipline.md rewritten — paired BAD/GOOD examples throughout,
the & escape rule, explicit \n vs <br/> ban, six-item self-check
before emit, simplify-rather-than-break ethos. Original ~250 tokens
expanded to ~440.
- latex-discipline.md new — Obsidian MathJax delimiters and \$ escaping
in prose to avoid accidental inline-math spans.
- concept-profile.md wired in {{include: latex-discipline}} alongside the
existing {{include: mermaid-discipline}}, placed immediately after the
"render a mermaid codefence here" instruction so the rules are in scope
at the moment of generation.
Streaming, timeout, and output-budget infrastructure (src/services/
directoryTemplateService.ts):
- streamPerplexityToFile now races each reader.read() against a fresh
per-chunk idle timer (270s deep-research, 90s normal — matches the
legacy modal flow at perplexityService.ts:659). Closes the
"Wall-clock-timeout cuts off long deep-research streams" issue.
- request-timeout-ms cft override is now the optional absolute wall-clock
ceiling (belt-and-suspenders backstop on top of the idle timer);
explicit 0 disables.
- max-tokens cft override added to buildPayload — passes through to
Perplexity's max_tokens parameter; lifts the silent ~8K-token default
ceiling that was clean-truncating thorough drafts via finish_reason:
length.
Documentation (README.md, docs/directory-templates.md):
- README — new top-level section "For Venture Capital, Private Equity, and
Equities-Trading Workflows" with a five-row table mapping each analyst
workflow to the template that produces it. Directory Templates section
updated to seven shipped templates plus the new per-template cft-block
knobs.
- docs/directory-templates.md — new "Per-template max-tokens override"
section with the diagnostic table for distinguishing wall-clock-timeout
truncation (mid-sentence cutoff, no sources footer) from max_tokens
truncation (clean section-end, full sources footer). The diagnostic
vocabulary that did not exist before this release.
Seeding (src/services/templateSeederService.ts):
- Registered all three new templates plus the new latex-discipline partial
so they auto-seed into vaults on first plugin load and appear in
Re-seed runs.
Version bump (manifest.json, package.json, versions.json):
- 0.2.1 to 0.3.0. minAppVersion unchanged at 1.8.10.
Engineering changelog entries (changelog/):
- 2026-05-26_02.md — the structural idle-timeout port.
- 2026-05-27_01.md — the two new templates, max-tokens override, and
rendering-discipline partial work. Includes the diagnostic-led
narrative on the vault-drift bug that caused the 4-of-5 broken-Mermaid
rate.
Release narratives:
- changelog/releases/0.3.0.md — full marketing-quality release narrative
with frontmatter, four-audience cascade, three diagnostic-led stories
(streaming-timeout, max_tokens, rendering-discipline), explicit upgrade
notes including the vault-drift warning, three deferred items for 0.4.x.
- release-notes/0.3.0.md — prose-only release-page body picked up by the
release.yml workflow when the 0.3.0 tag is pushed. Tighter cut of the
same narrative, suited for the GitHub release page.
Context-v stamp:
- context-v/issues/Wall-Clock-Timeout-Cuts-Off-Long-Deep-Research-Streams.md
marked status: resolved, date_resolved 2026-05-26, with the four open
items reorganized into "what landed" (3 of 4) and "deferred follow-ups"
(cross-service audit, settings-pane exposure, deep-research detection
robustness).
Files changed:
- manifest.json, package.json, versions.json
- README.md, docs/directory-templates.md
- src/services/directoryTemplateService.ts
- src/services/templateSeederService.ts
- src/docs/templates/README.md
- src/docs/templates/concept-profile.md
- src/docs/templates/market-map-profile.md
- src/docs/templates/market-category-profile.md (new)
- src/docs/templates/standards-and-specs-profile.md (new)
- src/docs/partials/mermaid-discipline.md
- src/docs/partials/latex-discipline.md (new)
- changelog/2026-05-26_02.md (new)
- changelog/2026-05-27_01.md (new)
- changelog/releases/0.3.0.md (new)
- release-notes/0.3.0.md (new)
- context-v/issues/Wall-Clock-Timeout-Cuts-Off-Long-Deep-Research-Streams.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
A new shipped template aimed at analyst-grade market maps (Known Category and Thesis-Driven flavors), and the timeout refactor that came out of running it against a real 7-8K-word draft and watching the wall-clock AbortController truncate the tail mid-sentence. The two ship together because the template surfaced the bug — running market-map-profile on lost-in-public/market-maps/Humanoid Robots and their Input Industries.md was the first observable instance of the wall-clock pathology — and they're both groundwork for the multi-stage Claude+Perplexity+RAG pipeline the exploration doc now sketches.
Impact: a fifth shipped template auto-seeded into Content-Dev/Templates/ on first plugin load (and re-seedable from settings), running on sonar-deep-research with return-images off by design; templates can override the plugin-level wall-clock timeout per-template via a new cft-block key; the plugin-level default jumped from 10 min to 30 min because the old default was tuned for short concept-profile runs and was cutting off any deep-research template that took its time. Existing four templates are unchanged in behavior — they don't declare the override and inherit the new 30-min default, which they all finish comfortably under.
Market-map template (src/docs/templates/market-map-profile.md):
- Heading skeleton: Market Snapshot, The Question this Map Answers, Why Now, Map of the Market, Lighthouse Examples, Innovator Profiles (Link/Offering/Funding/Why/Coverage cards + per-sub-segment summary table), Media-Voices-Coverage, Market Dynamics (Sizing, Adoption, Capital Flow), Frontier and Open Questions, Adjacent Concepts.
- Dual-flavor: the system prompt has the model pick between Known Category (Humanoid Robots, Quantum Computing) and Thesis-Driven (e.g. Neural Network Hardware as Brains for Robotics) based on title and tags.
- Anti-incumbent editorial stance copied from the concept template family — cap big tech at 1 of 5-10 in every sub-bucket, attribute innovation to startups/labs/founders.
- Declares request-timeout-ms: 2400000 (40 min) with an inline comment explaining the budget — the only shipped template that overrides the plugin default today.
- Registered in templateSeederService.ts so first-load seeding and the Re-seed button both pick it up; surfaced in src/docs/templates/README.md and docs/directory-templates.md.
Timeout refactor (src/services/directoryTemplateService.ts, main.ts):
- Per-template cft-block key request-timeout-ms: accepts number or numeric string, silently falls back on non-positive / non-numeric values, otherwise wins over the plugin-level setting for that template's runs. Resolution code sits just before the streamPerplexityToFile call so the rest of the streaming primitive is untouched.
- Plugin default directoryTemplatesRequestTimeoutMs bumped 600000 → 1800000 (10 min → 30 min). Settings-pane description rewritten to call out the override and the cost framing ("$10-$50 of analyst time per good output is worth waiting for").
- Structural fix (idle-timeout discipline ported from perplexityService.ts:659-668, where the legacy modal flow already does it correctly) is filed as its own open issue, not in scope for this commit.
Context-v writing:
- context-v/explorations/Multi-Stage-Cooperative-Claude-and-Perplexity-with-RAG.md — three architectures weighed (zoned single-file appends, multi-doc folder per market map, per-section cft-section blocks), an include-sources YAML schema proposed covering vault paths/globs and Chroma queries, and the "Perplexity and Claude do not overwrite each other until editing" heuristic translated into three implementation strictness levels. Recommends Option A (zoned appends) for v1.
- context-v/issues/Wall-Clock-Timeout-Cuts-Off-Long-Deep-Research-Streams.md — the Humanoid Robots truncation as motivating symptom, side-by-side diagnosis of the two timeout disciplines in the codebase (wall-clock here, per-chunk idle in the legacy modal), the partial fix shipped today, why it's only partial, the structural-fix proposal, and four pre-spec open items. Cross-linked both ways to the exploration so the issue is reachable from either entry point.
Files changed:
- src/docs/templates/market-map-profile.md (new)
- src/services/templateSeederService.ts
- src/services/directoryTemplateService.ts
- main.ts
- src/docs/templates/README.md
- docs/directory-templates.md
- context-v/explorations/Multi-Stage-Cooperative-Claude-and-Perplexity-with-RAG.md (new)
- context-v/issues/Wall-Clock-Timeout-Cuts-Off-Long-Deep-Research-Streams.md (new)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Records the 0.2.1 release commit hash in the release_commit frontmatter
field of changelog/releases/0.2.1.md, mirroring the stamp pass done for
0.2.0. The narrative now points at the exact commit the release was cut
from.
Files changed:
- changelog/releases/0.2.1.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Directory-template runs with sonar-deep-research were silently discarding
~99% of every response. Perplexity's deep-research model delivers its whole
finished document in the first SSE event's message.content; the streaming
reader took delta.content (correct for the token-streaming sonar models) and
captured only a ~75-character tail fragment of a ~10,000-character report.
toolkit-profile shipped with sonar-deep-research as its default model, so
the default was the broken one. This release fixes the streaming, hardens it
against mid-stream disconnects, and adds the tooling the directory-template
research workflow was missing.
Both streaming paths now read Perplexity's cumulative message.content
snapshot and append only the new tail — correct for token-streaming sonar
models and for deep research's first-event document dump alike. A timed-out
or dropped stream now keeps the partial content and the citations it already
received instead of discarding the whole run. "Apply directory template"
gained a run dialog with a per-run model selector, so the model is no longer
locked to the template file, and the loading notice names the model that is
actually running instead of a hardcoded "deep research" string.
Streaming (perplexityService.ts, directoryTemplateService.ts):
- Read message.content (cumulative snapshot), diff against written text,
append the tail; a startsWith guard skips a changed prefix rather than
corrupting the note.
- streamPerplexityToFile returns a partial result with a `truncated` flag on
a read error instead of throwing; applyTemplate writes the partial with
its sources footer and a truncation notice. The modal catch block salvages
its citation metadata the same way.
- Deep-research idle timeout 90s -> 270s; request-timeout default 5 -> 10 min.
Model selection (DirectoryTemplateRunModal.ts, main.ts):
- New run dialog: a template dropdown plus a Perplexity model dropdown that
defaults to the template's cft model and overrides it for that run only.
- cf_last_run_model now stamps the model that actually ran.
Search scoping (directoryTemplateService.ts):
- search_domain_filter support via cf_search_domains (target frontmatter)
and search-domains (cft block); allow/deny entries, capped at 10; job
boards are denied automatically.
Template (src/docs/templates/toolkit-profile.md):
- System prompt rebuilt: anchors the entity on its real name rather than the
scraped marketing tagline, makes source quality a triage task instead of
an allow/deny list, adds a search-don't-recall rule and per-section length
ceilings. Default model -> sonar-pro.
Files changed:
- src/services/perplexityService.ts
- src/services/directoryTemplateService.ts
- src/modals/DirectoryTemplateRunModal.ts
- main.ts
- src/docs/templates/toolkit-profile.md
- manifest.json
- package.json
- versions.json
- changelog/releases/0.2.1.md
- release-notes/0.2.1.md
Also included:
- manifest.json description aligned with package.json ("Perplexica (now Vane)").
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Workflow .github/workflows/release.yml looks for release-notes/<tag>.md
at the tagged commit; without this file it falls back to --generate-notes,
which produces a flat commit list instead of the marketing-shape body.
Mirrors the convention introduced in release-notes/0.1.2.md.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
README sections added/updated:
- Lede + Key Features list — Gemini in the provider list
- Network use table — new Gemini row + a note on the grounding-redirect resolution requests
- Initial Setup — new step 2 "Configure Google Gemini" between Perplexity and Perplexica; LM Studio renumbered to step 4
- New "Using Google Gemini" section between Using Perplexity and Using Perplexica, including why Gemini's per-segment citations beat Claude's
- Command Reference — new "Google Gemini Commands" subsection
- Directory Templates — new "Partials and preambles" subsection documenting the {{include: name}} directive + auto-attached preambles paradigm
- Table of Contents — Gemini link added
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Gemini lands as the fourth research provider — Google Search grounded,
per-segment citation attribution via groundingSupports[] (the layer Claude's
web_search_20260209 dynamic-filter sandbox loses), redirect URLs resolved
through Obsidian's requestUrl with canonical/og:url + <title> parsing so
citations land with real source URLs and real page titles.
Plus the partials + preambles paradigm (vault-visible shared rules across
the four directory templates), the addClass token bug that was silently
breaking eight settings sections finally fixed, system-prompt textareas
promoted from 200px right-edge slots to full-width 3-line rows, seeder
made idempotent at both folder and file layers, and ERR_NETWORK_CHANGED
mid-stream errors translated into plain-English re-run guidance.
Per-day changelogs:
- changelog/2026-05-19_01.md — partials + preambles + 11 hidden sections fix
- changelog/2026-05-19_02.md — Gemini provider lands
- changelog/2026-05-20_01.md — Obsidian-shipping pass (today)
Release narrative: changelog/releases/0.2.0.md
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two threads bundled because they were one session's work and the changelog tells the joined story:
1. PARTIALS + PREAMBLES — shared guidance moves out of TypeScript constants and out of duplicated template blocks into vault-visible markdown that templates pull in by name.
Three peer folders under templatesRoot's parent:
zz-cf-lib/templates/ (existing — your four profile templates)
zz-cf-lib/partials/ (new — pulled in via {{include: name}})
zz-cf-lib/preambles/ (new — auto-attached to every Perplexity request)
What moved:
- INLINE_CITATION_DIRECTIVE → src/docs/preambles/inline-citation.md
- IMAGE_PLACEMENT_DIRECTIVE → src/docs/preambles/image-placement.md
- buildResearchFraming() → src/docs/preambles/research-framing.md
- mermaid syntax discipline block (was duplicated across all four profile templates) → src/docs/partials/mermaid-discipline.md
New engine surface in src/services/directoryTemplateService.ts:
- expandIncludes(app, text, partialsRoot) — async, recursive, max depth 5, cycle-detected. Missing partials surface as inline [[include: <name> — file not found]] markers; cycles as [[include: <name> — cycle detected]].
- loadPreamble(app, preamblesRoot, name) — vault first, bundled default fallback with console.warn.
- parsePreambleOverrides(cftConfig) — per-template { system, skip-user, skip-all } in the cft fence.
- applyTemplate refactored: expand-then-interpolate for cftSystem and userSkeleton; preambles resolved (each expanded + interpolated) then assembled before sending.
DirectoryTemplateSettings extended with partialsRoot, preamblesRoot, systemPreambles, userPreambles. PerplexedPluginSettings + DEFAULT_SETTINGS mirror these (defaults: zz-cf-lib/partials, zz-cf-lib/preambles, [inline-citation], [{research-framing,always},{image-placement,return-images}]). Four new settings-tab rows expose them. buildDirectoryTemplateSettings() helper replaces two duplicated dirSettings literals.
templateSeederService.ts rewritten around a shared seedFolder() helper. Now seeds three folders (templates / partials / preambles) with the same idempotent rule (README always; content files only when target folder is missing or empty). Exports BUNDLED_PREAMBLES for runtime fallback when a vault preamble file is absent. seedTemplatesIfMissing and reSeedMissingFiles both take optional partialsRoot/preamblesRoot args.
The four bundled profile templates now reference {{include: mermaid-discipline}} in place of the previously-duplicated checklist block.
2. SETTINGS-UI FIX — eight call sites in main.ts had activeDocument.createEl('textarea'), an Obsidian global that's undefined in current Obsidian versions. The first one threw at the Perplexity "Request body template" textarea and aborted PerplexedSettingTab.display() before any of the other 11 sections (Claude, Perplexica/Vane, LM Studio, prompt blocks, Directory templates, Find Images) could mount. All eight switched to containerEl.createEl, matching every other DOM operation in the file. Settings tab now renders end-to-end.
Build green: pnpm build (eslint + tsc + esbuild production) exits 0.
Architecture review and design rationale captured at content-farm/context-v/issues/Partials-And-Preambles-For-Perplexed-Templates.md before implementing.
Vault-seeding caveat unchanged: existing vault templates won't auto-replace on plugin update. Copy the four updated bundled templates into your vault manually (or delete + reload) to pick up the {{include: …}} directive; the seeder will create zz-cf-lib/partials/ and zz-cf-lib/preambles/ on next plugin load.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three things bundled:
1. Release workflow at .github/workflows/release.yml fires on every 3-digit-semver tag push. Builds main.js + styles.css in a clean Ubuntu runner, signs all three release assets via actions/attest-build-provenance@v1 (sigstore-backed via OIDC), creates the GitHub Release with notes pulled from release-notes/<tag>.md. Identical to the workflow landing across cite-wide, image-gin, metafetch as a family-wide pattern.
2. release-notes/0.1.2.md authored in marketing-shape per content-farm/context-v/skills/changelog-conventions/SKILL.md "These are marketing artifacts" section.
3. package.json description aligned with manifest.json. Previous package.json text "A plugin for Obsidian that allows you to generate source-cited content..." still contained the bot-flagged "Obsidian" + self-reference patterns that we'd already fixed in manifest.json; this synchronises the two so anything reading from npm-style metadata gets the same active-voice description users see in the directory listing.
Version bumped 0.1.1 → 0.1.2 across manifest.json / package.json / versions.json. No source-code changes — Directory Templates, the four shipped templates, the first-run seeder, the Claude integration, the wide modals all behave identically to 0.1.1.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
styles.css is regenerated on every pnpm build / pnpm dev and minification differs run-to-run, so the tracked version was perpetually dirty in git status — same chronic noise pattern across cite-wide / image-gin / perplexed / metafetch. main.js was already gitignored for the same reason; this just brings styles.css onto the same footing.
Local devs still build styles.css for symlinked-vault testing. The GitHub Release attaches the freshly-built styles.css from the developer's filesystem at release time (gh release upload doesn't care about gitignore), so end users installing from the Community Plugins directory still receive it normally.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The fundingUrl is meant to be a tip-jar / sponsorship destination per the
Obsidian manifest spec — was previously set to https://lossless.group (the
group homepage, no way to send money). Corrected to
https://buymeacoffee.com/losslessgroup, which is the actual destination
across all Lossless-authored plugins.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
User-facing roll-up of the per-day changelogs covering the 0.1.0 → 0.1.1 span: Directory Templates paradigm (cf/cft codefence DSL, four shipped templates, auto-seeder, streaming + image-marker pipeline, anti-incumbent editorial stance), Claude as third research provider, redesigned wide modals, selection-anchored Find Images, cite-wide-compatible sources footer, and marketplace availability through the new community.obsidian.md portal.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Round 3 of the marketplace submission (#12513) flagged 73 errors + 6
warnings against the bot's recommended eslint config. This ships the
literal fixes the bot demanded so the rescan goes clean and the PR
moves to human review.
Sentence-case (66 errors). Applied the bot's exact "Expected: '...'"
rewrites across main.ts, the 8 modal files, and claudeService.ts. The
bot runs obsidianmd/ui/sentence-case with `enforceCamelCaseLower: true`
and no brand allowlist, so "Ask Perplexity" becomes "Ask perplexity",
"LM Studio" becomes "Lm studio", "PromptsService" becomes "promptsservice"
— UI text now reads as broken. Brand capitalization is to be restored
after marketplace acceptance, when we own the plugin entry on our own
update cadence. Local eslint.config.mjs's brand-allowlist override
also dropped so the local lint stops disagreeing with the bot.
Async without await (2 errors). main.ts:504 `callback: async () => {}`
reduced to `callback: () => {}` (the wrapped reinitializeServices() is
sync). directoryTemplateService.ts:195 listTemplates reduced from
`async function ... : Promise<TemplateFile[]>` to sync; the two callers
in main.ts updated to drop `await`.
no-restricted-globals (4 errors + 4 description-missing errors). All
four streaming fetch() sites (directoryTemplateService, lmStudioService,
perplexicaService, perplexityService) switched to `activeWindow.fetch`,
matching the precedent already used for `activeWindow.setTimeout`
elsewhere. The bare eslint-disable directives removed. SSE/streaming
rationale (Obsidian's requestUrl buffers) preserved in adjacent
comments.
Unused catch bindings (6 warnings). Three in main.ts, plus
lmStudioService:243, perplexicaService:255, logger.ts:47 — all
collapsed from `catch (e)` / `catch (error)` to `catch { ... }`.
Also included: changelog/2026-05-12_01.md documenting what the bot
flagged, the diff cost (brand names lowercased mid-string), the
non-sentence-case fixes, and verification.
Verification: both eslint configs (local + bot-mirror recommended)
exit 0; tsc -noEmit -skipLibCheck exit 0; production esbuild exit 0.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Backfills the documentation gap for the directory-templates feature
shipped in changelog/2026-05-10_01.md. Until now, a GitHub reader
landing on README.md had no idea the feature existed — the README
jumped straight from Command Reference into Developer Onboarding.
README.md — new "## Directory Templates" section between Command
Reference and Developer Onboarding (and a matching TOC entry).
~60 lines covering: why this exists, the three primitives
(template / commands / cleanup pipeline), the four shipped
templates and their target globs, auto-seed behavior including
the two-tier policy (README always, templates only when fresh),
how to write your own template, and a link to the canonical
reference.
docs/directory-templates.md — new canonical reference for repo
readers. Adapted and expanded from src/docs/templates/README.md
(which stays as the user-vault-seeded variant). Adds: why this
paradigm exists (the 1,600-empty-files problem), full anatomy
of the three template zones with grammar example, interpolation
token table, fill-vs-append run modes, streaming + cleanup
pipeline mechanics (think wrap, image marker swap, fallback
images section, sources footer aligned to cite-wide's
REFDEF_NUM_RE), frontmatter stamps (cf_last_run /
cf_last_run_model / google_books_url), the anti-incumbent
editorial stance rationale, first-run + re-seed semantics with
the two-tier policy, instructions for writing your own template,
plugin settings reference, and the open-items list mirrored from
the changelog. Links back to the changelog as the canonical
engineering record.
src/docs/templates/README.md left untouched — it's the
user-vault-seeded doc and remains correctly scoped for the
"you have these files in your vault, here's how to use them"
audience.
Naming note: the feature was originally specced as "prompt
outlines" (see context-v/specs/Using-Files-as-Prompt-Outlines.md).
The shipped naming is "directory templates" — matches the code,
the changelog, and the command-palette entries. The new doc
mentions the original name once, for orientation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ObsidianReviewBot's last scan (commit 4423052f, 2026-05-09) flagged
three `require-await` violations. The bodies of all three are fully
synchronous — promoting them to sync is the correct fix and removes
the false-positive shape for the marketplace scanner.
- main.ts#reinitializeServices: instantiates services synchronously
inside try/catch blocks. Drop `async`/`Promise<void>`; drop the
unnecessary `await` at the lone call site in the settings-saved
callback.
- claudeService.ts#afterMessage: walks message content and writes
to the editor via `replaceRange` — both sync. Drop `async`; drop
the `await` at the two call sites in the Claude streaming flow.
- perplexityService.ts#processStreamingMetadata: post-processes the
streaming response (think blocks, image placement, search-result
footer). All operations are sync editor mutations and string
processing. Drop `async`; drop the `await` at the lone call site.
`pnpm build` passes (eslint clean + tsc + esbuild). Local lint was
already clean against the obsidianmd recommended config — the bot
enforces `@typescript-eslint/require-await: "error"` while the
recommended config disables it, hence the divergence.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Captures the full arc of the directory-template paradigm: cf/cft codefence
DSL, four shipped templates (concept, vocabulary, source, toolkit), streaming
+ cleanup pipeline with image marker placement and fallback, anti-incumbent
editorial stance, cite-wide-compatible sources footer, frontmatter run-stamps
(cf_last_run, cf_last_run_model, google_books_url for books), and the
first-run seeder that drops shipped templates plus a README into the user's
vault on plugin load.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Original seeder skipped the entire write step when the folder already had
markdown — which meant a user with templates in place but no README never
got the README. The fix splits the policy in two:
- README: docs, always written if missing. Survives plugin updates and
developer-mode situations where the folder was hand-populated before
the seeder existed.
- Templates: user-customizable content. Only seeded when the folder is
fresh (no non-README markdown). A folder with even one template in it
is treated as user-managed and left alone, so a user who deleted
concept-profile intentionally doesn't get it resurrected.
folderHasMarkdown -> folderHasTemplate ignores README when deciding if a
folder is user-populated. Reason field gains a new "readme-only" state
for the case where the folder had templates but was missing the README.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
A freshly-installed perplexed has no templates folder, so the directory-template
commands have nothing to match. Users had to copy templates from the plugin
source by hand. Now the four shipped templates (concept-, vocabulary-, source-,
toolkit-profile) plus a README are inlined into main.js via esbuild's text
loader for .md files, and a seeder writes them into the configured templates
root on plugin load — but only if the folder is missing or contains no markdown,
so existing customizations are never touched.
A new "Re-seed templates" button in the Directory templates settings section
fills in any shipped file whose filename doesn't already exist (for pulling in
new templates after a plugin update).
Pieces:
- esbuild.config.mjs: register `.md` text loader so markdown imports work
- src/types/markdown.d.ts: TS shim for `*.md` imports
- src/services/templateSeederService.ts: seedTemplatesIfMissing +
reSeedMissingFiles (idempotent, never overwrites)
- src/docs/templates/README.md: end-user docs for the template system
(zone anatomy, interpolation tokens, commands, image markers, citation
behavior, frontmatter stamps, writing a custom template)
- main.ts: seeder fires after settings load; settings tab gains a Re-seed
button
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
For Sources/ entries the model identifies as books:
1. Template: if frontmatter already has google_books_url, treat as
authoritative and use it as the primary "Where it lives" link.
Otherwise, search Google Books and find the canonical URL.
2. Post-processor in directoryTemplateService: harvest the first
books.google.com or google.com/books/edition URL from generated
body and stamp it to frontmatter as google_books_url. Only stamps
when frontmatter doesn't already carry the field — user-curated
URLs are never overwritten.
This makes the URL findable in body content immediately after a run,
and self-canonicalizes into frontmatter so subsequent runs skip the
search. Same pattern as cf_last_run / cf_last_run_model: frontmatter
is the canonical home for canonical IDs.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The Sources/ directory holds heterogeneous trusted references — books,
people, podcasts/YouTube channels, magazines, newspapers, websites,
academic journals, journal networks, research reports, and events.
Previous source-profile was a generic three-section scaffold that didn't
match the heterogeneity.
New template takes one outline and adapts emphasis per type via three
mechanisms:
1. Type determination up front — system prompt names the seven canonical
types and tells the model to pick one from frontmatter signals
(`youtube_channel_url`, `wikipedia_url`, `aliases`, etc.) plus a
search.
2. Type-aware emphasis directives in the system prompt — explicit
per-type guidance for what each section should emphasize.
3. Conditional bullet shapes inside the user-prompt sections — Type and
Format, The People Behind It, and Catalog of Notable Works each list
per-type bullet shapes; the model picks the one that fits.
Sections: Identity & lede, Type and Format, The People Behind It, Catalog
of Notable Works, Why It Matters to Innovators, Best Starting Points,
Adjacent Sources.
Model: sonar-pro (matches concept- and vocabulary-profile). Source
preference biased toward the source's own primary surface over Wikipedia.
Image search note acknowledges that source images (covers, headshots,
logos, cover art) actually return well from Perplexity image search,
unlike abstract concept entries.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The article-generator path has worked for images all along because it has a
fallback: when the [IMAGE N: …] regex doesn't replace anything but Perplexity
returned images, it still surfaces them as a `## Images` section. The
directory-template flow had no such fallback — if the regex missed for any
reason (model emitted markers in an unexpected shape, model didn't emit
markers at all), the images dropped silently.
Three changes:
1. processContentWithImages now returns { content, replaced } so the caller
can branch on whether any markers actually got replaced.
2. New buildFallbackImagesSection emits a `# Images` block (heading +
embed + source line) appended just before the sources footer when
replaced === 0 && images.length > 0.
3. Permissive regex: also matches `[Image N: …]` (case-insensitive was
already there) plus the markdown-image-shaped `` and
`[IMAGE N](…)` forms that some models emit when they try to anticipate
the embed.
Diagnostics: console.debug logs the images.length / replaced count / model
on every run so the failure mode is observable in dev tools. Notice fires
when the fallback inserts images so the user knows what happened.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Concept-profile was using sonar-deep-research, which the article-generator's
existing compatibility-warning has long flagged as "Images are unstable in
deep research mode" — Perplexity returns an empty images array, so the
[IMAGE N: …] markers the model emits never get swapped for embeds and the
template's image-placeholder text leaks into the output.
Two changes:
1. Concept-profile drops to sonar-pro to match vocabulary-profile. Both are
encyclopedia-style entries that don't need the multi-step research loop;
sonar-pro is also the model that reliably returns images.
2. Post-stream cleanup now counts [IMAGE N: …] markers before and after
replacement. When markers remain unreplaced, log a console warning and
show a Notice telling the user how many markers stayed and how many
images Perplexity actually returned — so the failure mode is visible
instead of silent.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
cf_last_run and cf_last_run_model in frontmatter already capture run timestamp
and model. The "_Generated <ts> via <model>._" line in the # Sources block was
duplicating that data and adding visual noise to every generated file.
buildSourcesFooter now takes only the sources array; the parameters that fed
the provenance line are gone. Frontmatter stamping is unchanged.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Lifts the existing image-placement pattern from PerplexityService into the
directory-template streaming flow:
- Capture `images` from the SSE stream alongside `search_results`
- When the cft block sets `return-images: true`, append an image-placement
directive to the user prompt instructing the model to emit
[IMAGE N: <description>] markers (replacing any "Image embed placeholder"
bullets in the skeleton)
- Post-stream, swap each marker for  using the captured
images array; strip any unreplaced placeholder bullets
Concept-profile flips return-images to true to benefit from this. Vocabulary
already had it on. The "Find images for selection" command remains as a
manual fallback.
This is the same pattern Generate One-Page Article has been using. Avoids
new infrastructure (no URL-path ranking, no multimodal re-rank, no headless
screenshots) — see context-v/issues/Nudgeing-AI-Search-to-Return-Contextually-Appriate-Images.md.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Encyclopedia-entry style: tight definition + lede, disambiguation ordered by
relevance to innovation consulting, optional etymology when meaningful, adjacent
vocabulary with shade-of-meaning notes, usage-in-practice quotations, and common
misuses. Uses sonar-pro (faster + better image candidates than deep-research)
and inherits the anti-incumbent editorial stance from concept-profile.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Concept entries kept attributing innovations to tech giants (Microsoft, Google, etc.)
because training data over-represents them. Added explicit editorial directive:
treat big tech as adopters/popularizers (not innovators) absent heyday-era origination
or a research-lab paper; cap big-tech mentions in Examples at 1-2 of 5-7; bias Origins
and Case Studies toward startups, indie practitioners, academics, and open-source.
Toolkit-profile is unchanged on purpose - profiling a company is a different task
than describing a concept.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three landings:
1. New {{basename}} interpolation token. Resolves strictly to the target
file's basename without .md, never falling through to a frontmatter
`title` field. Concept-profile uses this so the H1 always matches the
filename in the vault, regardless of what the frontmatter says.
2. New seed template src/docs/templates/concept-profile.md (matched in the
user's vault at zz-cf-lib/templates/concept-profile.md). applies-to-paths
is "concepts/**". Heading skeleton:
- Defining and Describing {{basename}} (image placeholder bullet,
optional mermaid diagram, italic lede, 2–4 sentence paragraph)
- Uses in Context
- History of Use (Origins, Evolution)
- Best Real-World Examples
- Case Studies
search-recency is "year" rather than "month" since concepts evolve more
slowly than products.
3. In-body provenance line inside the # Sources footer:
_Generated <ISO timestamp> via <Provider> <model id>._
Always renders, even when the run returned zero sources (in which case
the body becomes "_No sources returned._"). Mirrors and complements the
frontmatter stamp; gives readers an at-a-glance record of which model
produced a given file and when, without needing to inspect frontmatter.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Every successful applyTemplate now writes two keys back to the target file's
frontmatter, via app.fileManager.processFrontMatter so other keys remain
untouched:
- cf_last_run: ISO 8601 timestamp (e.g. "2026-05-09T20:14:32.451Z").
- cf_last_run_model: provider + model id (e.g. "Perplexity sonar-deep-research").
Lets the vault be queried for staleness — e.g. "show all files in
Tooling/AI-Toolkit/Agentic AI/Agentic Workspaces sorted by oldest cf_last_run"
to find which entries need a refresh. Mirrors the og_last_fetch pattern from
metafetch.
Stamping happens only on { status: 'applied' } — errors and skips do not
touch frontmatter.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Reference section now emits as:
{content}
***
# Sources
[1]: [Title](URL)
[2]: [Title](URL)
Blank line before the separator, blank line between separator and the
h1 heading (Sources promoted from h2 to h1), blank line before the first
reference. Matches the user-preferred citation-block shape per the
Lossless citation spec aesthetic.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Previous version returned third-party stock illustrations (wufoo.com,
thingsup.io, pixelfreestudio.com) instead of actual screenshots from the
entity's site. The "prefer entity domain" instruction in the prompt was
suggestion-only; the model picked whatever Perplexity's image search
indexed against the topic keywords.
Two changes that should compose to fix this:
1. Send `search_domain_filter: [entityDomain]` on the Perplexity request
when the active file's frontmatter has a `url` field. This restricts
Perplexity's underlying web search to the entity's domain, so the model
can only see and reference on-site pages.
2. Strict client-side filter as defense in depth. Any returned image whose
`origin_url` or `image_url` hostname does not end in the entity domain
is rejected before placement. Better to surface "no on-domain images
found" via Notice than to silently insert third-party stock images.
Prompt also tightened: it now instructs the model that searches are
already domain-filtered and that returning fewer images (or none) is
preferred over substituting generic stock content.
Next escalation if this still falls short: direct DOM crawl of the entity
URL (fetch HTML, parse <img> tags by alt-text/context match) — fully
deterministic, no model judgment for image selection.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
New command "Find images for selection". Highlight a passage in any open file,
invoke the command, and the plugin:
1. Splits the selection into paragraphs and numbers them.
2. Reads the active file's frontmatter for url / site_name (entity context).
3. Sends a Perplexity request with return_images=true asking the model to:
- Find N screenshot/product/feature images that illustrate the specific
passage content.
- Prefer images hosted on the entity's domain; fall back to other web
sources only when on-domain images aren't available.
- For each chosen image, output one line of the form
`[AFTER {paragraph_number}] [IMAGE {n}: <description>]` so the model
itself decides which paragraph each image belongs after.
4. Parses the placement markers, maps each image number to the corresponding
entry in response.images, and rebuilds the selection with `\n\n`
separators around each embed so images sit between paragraphs, never
inside one.
5. Falls back to even-distribution domain-preferred ordering if the model
emitted no placement markers.
Result is `` embeds in standard CommonMark/GFM form,
inserted via editor.replaceSelection. Frontmatter and surrounding content
are untouched.
New setting `findImagesMaxImages` (default 3) caps the per-invocation count.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Previous fix used [N] without the colon. Lossless-Citation-Spec.md mandates
the colon for reference-section identifiers (cite-wide reminders/spec line
68: "always use a `: ` after the citation identifier"). Cite-wide's parser
accepts both shapes, but the colon form is canonical and what other
Lossless tooling expects to see in checked-in files.
Verified the parser flow against cite-wide:
- REFDEF_NUM_RE (llmCitationParserService.ts:91) matches both `[N]` and `[N]:`
- reformatRefDefBodyAsMarkdownLink (line 547) skips wrapping when the body
already contains a markdown link, so `[N]: [Title](URL)` passes through
untouched into the hex-conversion stage.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The previous footer used ordered-list style ("1. [Title](URL)"), which
cite-wide's LLM citation parser ignores. Cite-wide's REFDEF_NUM_RE expects
each line to start with [N] (square-bracketed digit), optionally followed by
a colon, then the link.
New format per source:
[1] [Title](URL)
[2] [Title](URL)
This matches cite-wide/src/services/llmCitationParserService.ts:91 verbatim,
so running cite-wide's "convert numeric citations to hex" command on a file
generated by this plugin now picks up the footer references and rewrites
them to the project's [^hexId] form.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Real-vault use revealed Perplexity Deep Research was returning all-essay
content padded from training data instead of search results: zero inline
citations, "not available in current accessible sources" repeated for every
section. Diagnosis traced two root causes — laziness (the system prompt's
escape hatch "even if a section reads 'limited public information'" gave the
model permission to dodge search) and hallucination (no enforcement that
factual claims be tied to search results).
Fixes in this commit:
1. Streaming writes (replaces v0.2's buffer-then-write). fetch() with SSE,
tokens flush to file every 500ms via app.vault.modify. Cancel mid-stream
via existing batch-cancel flag; partial output remains in file.
2. Combined system prompt: perplexed's standard inline-citation directive
prepends the template's role framing, matching how the working ask-
perplexity command activates search.
3. Auto-prepended research framing to user prompt: "Research the entity X
using web search... Every factual claim must be followed by [N]." The
prompt no longer reads as a writing brief.
4. Post-stream cleanup runs after the write completes (per user spec):
- Wraps any <think>...</think> blocks in ```think-output fenced code (raw
<think> HTML breaks Obsidian rendering).
- Appends a ## Sources footer listing search_results from the API
response so evidence-of-search is visible even if the model failed to
inline [N] markers.
5. Seed template (toolkit-profile.md) system prompt rewritten with positive
directives only — no escape hatches, no "limited info" phrasing, no
failure-mode descriptions. Encodes the project rule that telling a model
it can be lazy makes it lazy.
ApplyOutcome now carries sourceCount; the user-facing notice reports it.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three additions driven by real-vault use against ~1600 Tooling/ profile files:
1. Folder batch run. New command "Apply directory template to all files in
folder" prompts for folder, then template, then a confirm modal showing
total/fill/append counts before firing. Sequential execution with a
live-updating progress notice. New "Stop directory template batch"
command sets a cancel flag checked between files.
2. Append-below replaces abort-on-existing-body. If the target body is
non-empty, the response is appended after a blank-line separator instead
of skipping. Frontmatter still byte-identical. ApplyOutcome return type
distinguishes fill / append / skip / error so batch can tally them.
3. Interpolation extended beyond {{title}} and {{frontmatter}}. Now supports
{{frontmatter.X}} for explicit access and bare {{X}} as auto-fallback to
frontmatter.X — so templates can use {{site_name}}, {{slug}}, etc.
directly. Arrays render as comma-joined strings; nested objects as YAML.
Also adds {{today}} → ISO date.
New files:
- src/modals/FolderPickerModal.ts — FuzzySuggestModal<TFolder>.
- src/modals/BatchConfirmModal.ts — confirm dialog showing batch counts.
Out of scope, still:
- v0.3: cf codefence in user notes.
- v0.4: section-aware writing + per-section output modes.
- v0.5: citation/backlink preservation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Match the user's preferred naming. The template file is "Toolkit Profile"
(its identity); it applies to entries under Tooling/ via applies-to-paths.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Solves the immediate workload of populating ~1600 nearly-empty profile files
under Tooling/, Sources/, and Vocabulary/ with a consistent outline + body.
The user picks a template from a fuzzy palette (filtered by glob match against
the active file's path), and the plugin sends a single Perplexity Deep Research
request driven by the template's heading skeleton + per-section bullet
instructions. The response is buffered, then written into the empty body of the
target file. Frontmatter is byte-identical before and after.
Template anatomy (per context-v/specs/Per-Directory-Profile-Templates.md):
- Frontmatter — title, applies-to-paths globs, description.
- Pre-`cft` prose — explainer for humans, ignored at runtime.
- One `cft` codefence — YAML config + `system:` field. Config-only; no prompt
body inside the fence.
- Post-`cft` skeleton — heading outline with model-facing bullet instructions.
Sent as the user prompt with `{{title}}` and `{{frontmatter}}` interpolated.
- A `***` line below the skeleton terminates the user prompt; everything below
is authoring scratch and never reaches the API.
New files:
- src/services/directoryTemplateService.ts — template loader, parser, glob
matcher, frontmatter whitelist filter, prompt assembly, non-streaming
Perplexity call, file writer.
- src/modals/DirectoryTemplatePickerModal.ts — FuzzySuggestModal wrapping the
glob-matched template list.
- src/docs/templates/{tooling-profile,source-profile}.md — seed templates
users can copy into their vault.
main.ts changes:
- New settings: directoryTemplatesRoot (default "zz-cf-lib/templates"),
directoryTemplatesFrontmatterWhitelist (default ["title", "og_description",
"tags", "og_image"]), directoryTemplatesRequestTimeoutMs (default 300000).
- New command: "Apply directory template to current file".
- New settings UI section under "Directory templates".
Out of scope for v0.1 — captured in the spec for later iterations:
- Folder-batch run with concurrency/progress/resume (v0.2).
- `cf` codefence runtime in user notes (v0.3).
- Section-aware writing + output modes (v0.4).
- Citation/backlink preservation (v0.5).
- Multi-provider, image generation, verb registry split (later).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Patch release bundling the marketplace-compliance lint pass for the
Obsidian community plugin re-submission. No user-visible behavior
changes from 0.1.0; this is the bot-clean version.
versions.json picks up "0.1.1": "1.8.10" matching the manifest's
minAppVersion.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Applies all required fixes from the bot's review of PR #12513
(obsidianmd/obsidian-releases). ESLint now passes 0 errors with the
Obsidian community ruleset; tsc passes; production build is clean.
Lint setup:
- Adopted eslint-plugin-obsidianmd@^0.2.9 in eslint.config.mjs
- Mirrored the bot's rule surface locally so violations surface in
`pnpm lint` instead of the marketplace PR thread
- Configured the ui/sentence-case rule with a brand allowlist
(Perplexity, Perplexica, Vane, Claude, Anthropic, LM Studio, Imgur,
ImageKit, OpenAI, Ollama, Sonar, Llama, GPT, YAML, JSON, URL, API)
so legitimate proper nouns aren't lowercased
Code fixes (487 → 0):
- console.log/info → console.debug across all sources (~130 sites)
- UI strings normalized to sentence case (~150 sites)
- Command IDs and names cleaned up: dropped "command" suffix, dropped
"perplexed" plugin-name prefix
- Settings tab section headers switched from createEl('h2') to
new Setting().setHeading() (5 sites)
- Inline element.style.color/width/minHeight/fontFamily migrated to
a new CSS class (.perplexed-json-textarea + .is-tall / .is-extra-tall)
in src/styles/settings-tab.css (8 textarea sites, 32 style assignments)
- Async input handlers wrapped: addEventListener('input', () => void
(async () => { ... })()) so the listener type matches (8 sites)
- forEach((opt) => dd.addOption(...)) blocks made void-returning to
satisfy no-misused-promises (9 modal sites)
- JSON.parse results typed as unknown then narrowed
- throw <string> → throw new Error(<string>)
- ${unknown} interpolations narrowed via instanceof Error
- Removed dotenv runtime dependency: published plugins shouldn't read
.env at runtime; user enters API keys via the settings tab
- Replaced builtin-modules dev-dependency with node:module's
builtinModules — same data, no extra package
- Logger console-method dispatch rewritten as a switch instead of
dynamic console[level] indexing (which the bot rejects)
Streaming exceptions:
- src/services/{perplexityService,lmStudioService,perplexicaService}.ts
retain `fetch()` for SSE / chunked streaming because Obsidian's
`requestUrl` buffers the whole body. Each site has an
`eslint-disable-next-line no-restricted-globals` with the marketplace
`/skip` justification inline. Plan to surface these on the PR with
a `/skip` reply.
Reference docs:
- context-v/issues/Obsidian-Review-Bot-Feedback-on-Perplexed-Submission.md
(issue log distilled into…)
- context-v/reminders/Obsidian-Marketplace-Compliance.md (the rules
themselves, reusable for image-gin and cite-wide submissions)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Perplexity's Sonar API stopped embedding inline citation markers in
message.content; responses now return prose plus separate citations
and search_results arrays. The plugin's addCitations() rendered a
correct bibliography, but the body had nothing to anchor the numbers
to.
Prepend a system message telling the model to append [1], [2], ...
markers inline, ordered to match search_results. All three payload
build sites now share one messagesWithSystem array.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Brings the manifest in line with the rest of the content-farm plugins —
1.8.10 is the team's current minAppVersion baseline (image-gin, plunk-it,
lmstud-yo, filestarter, file-transporter all use it); the 0.15.0 here
was a 2022-era inheritance from the original Obsidian sample plugin.
Also adds fundingUrl for parity with the rest of the farm.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Tighten the retrospective to lead with this repo's own substance.
Lineage from the official Obsidian sample plugin is what every plugin
forks from on day one — not interesting and now relegated to a footnote
or removed entirely.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Seeds the changelog/ directory with a retrospective entry dated to the
first commit of substance in this repo, written in May 2026 from the
content-farm pseudomonorepo cleanup. The convention didn't exist when
the original work shipped; this entry exists so the upcoming GitHub
Content API roll-up loader has something from this plugin to surface
on the content-farm splash.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds a context-v/reminders/ note that points future humans and AI agents
to the canonical Obsidian API source of truth and lists the most common
community plugin review-bot rejection reasons. Same canonical reminder is
being seeded across every plugin in the content-farm ecosystem so the
content-farm splash's rolled-up context-v feed has reminders surfacing
from every plugin once the GitHub Content API loader lands.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds the Updating-Naming-to-Reflect-Provider-Rebrand issue note that
documents the surfaces touched by the Perplexica/Vane rename and the
ones intentionally left as internal symbols.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
