@steipete/summarize 0.11.1 → 0.13.0

package/CHANGELOG.md

@@ -1,10 +1,82 @@
 # Changelog
 
-## 0.11.1 - 2026-02-14
+## 0.13.0 - UNRELEASED
+
+### Features
+
+- Slides: support `--slides` for local video files in the main CLI and `summarize slides`, route local videos through the shared slide-aware flow, and document the local-file workflow (#149, thanks @steipete).
+- Models: add explicit `github-copilot/...` model support backed by GitHub Models, including shorthand ids like `github-copilot/gpt-5.4` and `GITHUB_TOKEN` / `GH_TOKEN` auth.
+- Models: add OpenCode as a first-class CLI provider across CLI flags, config, auto fallback, daemon picker/chat flows, and Chrome extension settings, while preserving existing OpenClaw behavior (#169, thanks @maciej).
+- CLI providers: add OpenClaw as a configurable CLI backend (`--cli openclaw`, `cli/openclaw/...`, `openclaw/...`) across config, daemon discovery, and docs (#165, thanks @yqf-ai).
+- Config: allow setting a default summary length via `output.length`, and keep prompt-override runs aligned with the configured length/language defaults in both CLI and daemon flows (#178, thanks @maciej).
+- Media detection/cache: recognize `.m3u8` HLS playlists as direct media inputs and preserve the playlist extension in the media cache (#159, thanks @mdsakalu).
+
+### Fixes
+
+- OpenAI models: route GPT-5.4 / GPT-5.4 mini / GPT-5.4 nano / GPT-5 mini / GPT-5 nano text requests through direct provider APIs instead of the stale generic parser, preserve the real `gpt-5.4-mini` / `gpt-5.4-nano` ids end-to-end, and fall GitHub Models OpenAI GPT-5-family requests back to `gpt-5-chat` when GitHub rejects the raw id.
+- GitHub Models: make `github-copilot/...` shorthand inference family-based instead of pinning old exact prefixes, so newer ids like `gpt-5.4`, `o5`, and `claude-opus-4.6` normalize correctly when the backend exposes them.
+- Slides/local video: transcribe direct videos for slide summaries, avoid fake local-file “Downloading audio” phases, and keep progress text visible while slide extraction runs.
+- Chrome extension slides: restore slide text/session state more reliably so reruns and reloads do not leave stale or blank slide summaries.
+- Transcription: retry Groq Whisper uploads via `curl` when Node multipart uploads get a 403, fixing local `.ogg` regressions on some environments.
+- YouTube: detect obviously truncated caption-track transcripts on long videos and fall through to yt-dlp transcription instead of caching a broken partial result (#184, thanks @sportiz91).
+- YouTube: treat yt-dlp “no audio stream” videos as a non-fatal unavailable transcript case so summarize can continue cleanly with an explanatory note (#161, thanks @mdsakalu).
+- Cache: include the prompt `<context>` block in summary cache hashing and bump the cache format version so stale cross-page summary collisions cannot be reused (#171, thanks @mvance).
+- CLI providers: stream OpenClaw prompts over stdin instead of `--message`, make daemon side-panel chat honor `openai.useChatCompletions`/custom OpenAI-compatible base URLs, and stop leaking raw Codex JSONL events like `thread.started` when no assistant text was produced.
+- Chrome extension: add a copy button for rendered summaries so results can be copied without manual selection.
+- Chrome extension chat: handle plain-string assistant replies in the side-panel agent loop instead of crashing on `.filter()` tool-call extraction (#186, thanks @Youpen-y).
+- Windows containers: let `summarize daemon install` start the daemon for the current container session without Scheduled Task registration, keep `0.0.0.0` binding Windows-only, and probe slide tools by spawning commands when PATH lookup is unreliable (#152, thanks @mathicg).
+- Windows daemon: keep Scheduled Task startup hidden without breaking `summarize daemon restart` or uninstall by tracking the hidden daemon PID and killing that process tree before reruns/removal (#146, thanks @mathicg).
+- Whisper.cpp: honor config-resolved transcription env overrides for readiness checks, model display, and local transcription so custom binary/model paths work outside `process.env` (#160, thanks @mdsakalu).
+- Daemon models: gracefully fall back for unrecognized custom models when using proxy base URLs instead of crashing on undefined API metadata (#175, thanks @douo).
+- Docs/setup: switch Homebrew instructions from the old tap to the official `brew install summarize` formula, including the side-panel setup UI and release checklist (#172, thanks @zeldrisho).
+- Chrome extension: detect blank `userAgentData.platform` browsers like Vivaldi by falling back to `navigator.platform` before choosing OS-specific setup instructions (#158, thanks @bytrangle).
+- Firecrawl: reject `--firecrawl always` for YouTube URLs with an explicit guidance error instead of silently skipping Firecrawl on the transcript-first path (#145, thanks @steipete).
+- YouTube: keep Gemini-only no-caption runs on the transcription path by forwarding the Google API key from the top-level URL flow into link-preview transcription config (#148, thanks @bytrangle).
+- Homebrew: make the tap formula fail clearly on Linux instead of installing a macOS binary, and add generator/test coverage for the macOS-only guard (#147, thanks @steipete).
+- Maintenance: update the GitHub Pages workflow to `actions/configure-pages@v6` and `actions/deploy-pages@v5` (#182, thanks @dependabot).
+
+## 0.12.0 - 2026-03-11
+
+### Features
+
+- Models: add `nvidia/...` provider alias (uses `NVIDIA_API_KEY` + optional `NVIDIA_BASE_URL`) for NVIDIA OpenAI-compatible endpoints.
 
 ### Fixes
 
+- Transcription: add AssemblyAI as a first-class remote provider across direct media, podcast/RSS, and yt-dlp YouTube fallback; refactor remote fallback ordering, expand config/env support (`ASSEMBLYAI_API_KEY`, legacy `apiKeys.assemblyai`), and add AssemblyAI unit + live coverage (#126).
+- X/Twitter: prefer `xurl` for tweet extraction when installed, fall back to `bird`, preserve long-form/article text plus media URLs, add live `xurl` extraction/media coverage, and replace the stale dead-`bird` install tip with a current X CLI recommendation (#70).
+- Models: make daemon agent `artifacts` schemas Gemini-safe, improve Google empty-response handling with preview-to-stable fallback, and switch CLI/auto Gemini defaults away from brittle preview behavior (#82, #96).
+- Agents: expand model auto-resolution errors with checked models, missing env/CLI setup, and daemon restart guidance (#107).
+- Daemon: support multiple saved extension tokens, migrate legacy single-token configs, and accept any configured token for auth (#116).
+- Chrome extension: harden side-panel slides so SSE keepalives no longer false-time out, seeded placeholders no longer block pending/cached slide runs, retries can start a fresh summarize+slides run, and reruns replace stale slide state.
+- Chrome extension: refactor side-panel navigation/run attachment policy so late summary/slide runs no longer attach to the wrong page after tab or URL switches, and expand headless regression coverage for pending-run resume and slide-mode transitions.
+- Chrome extension: default fresh installs to slide mode, keep passive tab navigation out of chat, and align slide cards with CLI `--slides` by preferring per-slide summary text over raw transcript/OCR fallback.
+- Chrome extension tests: add stronger YouTube slide E2E coverage for loaded images, summary-backed slide text, and switching between videos mid-analysis without stale slide-summary bleed.
+- Chrome extension: isolate slide-summary stream callbacks per run and harden Playwright settings hydration so late events no longer blank slide text when switching videos mid-analysis.
+- Transcription: add Gemini audio/video transcription support across direct media, podcast/RSS, and yt-dlp YouTube fallback, including Files API uploads for larger media plus new Gemini live coverage (#89).
 - npm packaging: publish CLI with `pnpm publish` so `@steipete/summarize-core` is version-pinned in published metadata (no `workspace:*` in registry package).
+- Slides: detect WezTerm as an iTerm-compatible terminal for inline slide images in `--slides` mode. (#133) — thanks @doodaaatimmy-creator.
+- CLI help: surface `summarize refresh-free` in `summarize help` output.
+- CLI: report CLI provider timeouts explicitly, including the duration, command, and a `--timeout` hint instead of collapsing them into generic exec failures (#100, thanks @christophsturm).
+- Daemon: restrict CORS responses to trusted extension and localhost origins, with regression coverage for allowed and denied `Origin` headers (#108, thanks @sebastiondev).
+- Transcription: chunk oversized Groq Whisper uploads with ffmpeg in file mode instead of failing out on files above the 30MB limit (#134, thanks @WinnCook).
+- Docs: tighten landing-page mobile layout so hero, cards, code blocks, and nav stay readable on narrow screens (#118, thanks @Acidias).
+- Release: build macOS x64 Bun artifacts and add regression coverage for Homebrew formula rewrites during dual-arch releases (#122, thanks @androidshu).
+- YouTube: tighten hostname validation across core, slides, and extension helpers so attacker-controlled lookalike hosts are no longer treated as YouTube URLs (#91, thanks @RinZ27).
+- Config: honor `zai.baseUrl` config fallback for blank env values and keep Z.AI base URL overrides working outside the summary flow (#102, thanks @liuy).
+- Chrome extension: tighten options and sidepanel UI spacing, copy actions, and advanced-controls layout for a cleaner panel experience (#86, thanks @morozRed).
+- Slides: warn in summary mode when `--slides` dependencies are missing, and document required local installs for `ffmpeg`, `yt-dlp`, and optional `tesseract`.
+- Docs: fix broken docs index links by setting an empty Jekyll `baseurl` (#113, thanks @Youpen-y).
+- Models: preserve model id casing after the provider prefix so OpenAI-compatible proxies can route exact names correctly (#128, thanks @WinnCook).
+- Cache: give extract entries with unavailable transcripts the same short retry TTL as negative transcript cache entries, so transient Apify failures can recover (#115, thanks @gluneau).
+- Daemon: apply the saved env snapshot to `process.env` before `daemon run` starts so child tools inherit the right PATH and API/tool config under launchd/systemd (#99, thanks @heyalchang).
+- Chrome automation: require sidepanel arming before debugger-backed native input can run in a tab, and auto-disarm after browser JS execution ends (#129, thanks @omnicoder9).
+- Media setup: fix the local whisper.cpp install hint to use the current Homebrew formula name `whisper-cpp` (#92, thanks @zerone0x).
+- CLI output: cap markdown render width on very wide terminals by default, with a `--width` override for manual control (#119, thanks @howardpen9).
+- Slides: size inline slide images from terminal width instead of keeping them pinned to 32 columns, capped at 2x the previous width while preserving `COLUMNS` fallback behavior (#125, #135, thanks @WinnCook).
+- Shell completions: add Fish shell completions for the current CLI flags and option values (#95, thanks @fbehrens).
+- Bun fetch: only opt into compressed HTML/YouTube responses when running under Bun, and retry link-preview fetches with `Accept-Encoding: identity` after Bun decompression failures (#105, thanks @maciej).
+- Daemon: support `cli/...` models in chat and agent endpoints, including CLI auto-fallback when no API-key transport is available (#109, thanks @jetm).
 
 ## 0.11.0 - 2026-02-14
 

package/README.md

@@ -1,26 +1,22 @@
 # Summarize 📝 — Chrome Side Panel + CLI
 
-![GitHub Repo Banner](https://ghrb.waren.build/banner?header=Summarize%F0%9F%93%9D&subheader=Chrome+Side+Panel+%2B+CLI&bg=f3f4f6&color=1f2937&support=true)
-
-<!-- Created with GitHub Repo Banner by Waren Gonzaga: https://ghrb.waren.build -->
-
 Fast summaries from URLs, files, and media. Works in the terminal, a Chrome Side Panel and Firefox Sidebar.
 
-**0.11.0 preview (unreleased):** this README reflects the upcoming release.
-
-## 0.11.0 preview highlights (most interesting first)
+## Highlights
 
 - Chrome Side Panel **chat** (streaming agent + history) inside the sidebar.
-- **YouTube slides**: screenshots + OCR + transcript cards, timestamped seek, OCR/Transcript toggle.
+- **Video slides**: screenshots + OCR + transcript cards for YouTube, direct video URLs, and local video files.
 - Media-aware summaries: auto‑detect video/audio vs page content.
+- Coding CLI backends: Codex, Claude, Gemini, Cursor Agent, OpenClaw, OpenCode.
 - Streaming Markdown + metrics + cache‑aware status.
 - CLI supports URLs, files, podcasts, YouTube, audio/video, PDFs.
 
 ## Feature overview
 
 - URLs, files, and media: web pages, PDFs, images, audio/video, YouTube, podcasts, RSS.
-- Slide extraction for video sources (YouTube/direct media) with OCR + timestamped cards.
-- Transcript-first media flow: published transcripts when available, Whisper fallback when not.
+- Slide extraction for video sources (YouTube, direct video URLs, local video files) with OCR + timestamped cards.
+- Transcript-first media flow: published transcripts when available, then Groq/ONNX/whisper.cpp/AssemblyAI/Gemini/OpenAI/FAL transcription fallback when not.
+- Coding CLI providers: Claude, Codex, Gemini, Cursor Agent, OpenClaw, OpenCode.
 - Streaming output with Markdown rendering, metrics, and cache-aware status.
 - Local, paid, and free models: OpenAI‑compatible local endpoints, paid providers, plus an OpenRouter free preset.
 - Output modes: Markdown/text, JSON diagnostics, extract-only, metrics, timing, and cost estimates.
@@ -42,7 +38,7 @@ YouTube slide screenshots (from the browser):
 
 1. Install the CLI (choose one):
    - **npm** (cross‑platform): `npm i -g @steipete/summarize`
-   - **Homebrew** (macOS arm64): `brew install steipete/tap/summarize`
+   - **Homebrew** (Homebrew/core): `brew install summarize`
 2. Install the extension (Chrome Web Store link above) and open the Side Panel.
 3. The panel shows a token + install command. Run it in Terminal:
    - `summarize daemon install --token <TOKEN>`
@@ -58,8 +54,9 @@ Notes:
 
 - Summarization only runs when the Side Panel is open.
 - Auto mode summarizes on navigation (incl. SPAs); otherwise use the button.
-- Daemon is localhost-only and requires a shared token.
+- Daemon is localhost-only and requires a shared token; rerunning `summarize daemon install --token <TOKEN>` adds another paired browser token instead of invalidating the old one.
 - Autostart: macOS (launchd), Linux (systemd user), Windows (Scheduled Task).
+- Windows containers: `summarize daemon install` starts the daemon for the current container session but does not register a Scheduled Task. Run it each time the container starts or add that command to your container startup, and publish port `8787` so the host browser can reach the daemon.
 - Tip: configure `free` via `summarize refresh-free` (needs `OPENROUTER_API_KEY`). Add `--set-default` to set model=`free`.
 
 More:
@@ -118,13 +115,37 @@ npm i @steipete/summarize-core
 import { createLinkPreviewClient } from "@steipete/summarize-core/content";
 ```
 
-- Homebrew (custom tap):
+- Homebrew:
+
+```bash
+brew install summarize
+```
+
+Homebrew ships from `homebrew/core` via `brew install summarize`.
+If Homebrew is unavailable in your environment, use the npm global install above.
+
+### Optional local dependencies
+
+Install these if you want media-heavy features:
+
+- `ffmpeg`: required for `--slides` and many local media/transcription flows
+- `yt-dlp`: required for YouTube slide extraction and some remote media flows
+- `tesseract`: optional OCR for `--slides-ocr`
+- Optional cloud transcription providers:
+  - `GROQ_API_KEY`
+  - `ASSEMBLYAI_API_KEY`
+  - `GEMINI_API_KEY` / `GOOGLE_GENERATIVE_AI_API_KEY` / `GOOGLE_API_KEY`
+  - `OPENAI_API_KEY`
+  - `FAL_KEY`
+
+macOS (Homebrew):
 
 ```bash
-brew install steipete/tap/summarize
+brew install ffmpeg yt-dlp
+brew install tesseract # optional, for --slides-ocr
 ```
 
-Apple Silicon only (arm64).
+If `--slides` is enabled and these tools are missing, Summarize warns and continues without slides.
 
 ### CLI vs extension
 
@@ -142,8 +163,8 @@ summarize "https://example.com"
 URLs or local paths:
 
 ```bash
-summarize "/path/to/file.pdf" --model google/gemini-3-flash-preview
-summarize "https://example.com/report.pdf" --model google/gemini-3-flash-preview
+summarize "/path/to/file.pdf" --model google/gemini-3-flash
+summarize "https://example.com/report.pdf" --model google/gemini-3-flash
 summarize "/path/to/audio.mp3"
 summarize "/path/to/video.mp4"
 ```
@@ -189,10 +210,18 @@ Spotify episode page (best-effort; may fail for exclusives):
 summarize "https://open.spotify.com/episode/5auotqWAXhhKyb9ymCuBJY"
 ```
 
+HLS playlist:
+
+```bash
+summarize "https://example.com/master.m3u8"
+```
+
 ### Output length
 
 `--length` controls how much output we ask for (guideline), not a hard cap.
 
+Set a default in `~/.summarize/config.json` with `output.length`.
+
 ```bash
 summarize "https://example.com" --length long
 summarize "https://example.com" --length 20k
@@ -235,14 +264,20 @@ Use gateway-style ids: `<provider>/<model>`.
 
 Examples:
 
+- `openai/gpt-5.4`
+- `openai/gpt-5.4-mini`
+- `openai/gpt-5.4-nano`
 - `openai/gpt-5-mini`
+- `openai/gpt-5-nano`
+- `github-copilot/gpt-5.4`
 - `anthropic/claude-sonnet-4-5`
 - `xai/grok-4-fast-non-reasoning`
-- `google/gemini-3-flash-preview`
+- `google/gemini-3-flash`
 - `zai/glm-4.7`
 - `openrouter/openai/gpt-5-mini` (force OpenRouter)
 
 Note: some models/providers do not support streaming or certain file media types. When that happens, the CLI prints a friendly error (or auto-disables streaming for that model when supported by the provider).
+`gpt-5.4-mini` and `gpt-5.4-nano` are treated as real model ids; the same shorthand also works under `github-copilot/...`.
 
 ### Limits
 
@@ -265,7 +300,7 @@ Use `summarize --help` or `summarize help` for the full help text.
 - `--length short|medium|long|xl|xxl|s|m|l|<chars>`
 - `--language, --lang <language>`: output language (`auto` = match source)
 - `--max-output-tokens <count>`: hard cap for LLM output tokens
-- `--cli [provider]`: use a CLI provider (`--model cli/<provider>`). Supports `claude`, `gemini`, `codex`, `agent`. If omitted, uses auto selection with CLI enabled.
+- `--cli [provider]`: use a CLI provider (`--model cli/<provider>`). Supports `claude`, `gemini`, `codex`, `agent`, `openclaw`, `opencode`. If omitted, uses auto selection with CLI enabled.
 - `--stream auto|on|off`: stream LLM output (`auto` = TTY only; disabled in `--json` mode)
 - `--plain`: keep raw output (no ANSI/OSC Markdown rendering)
 - `--no-color`: disable ANSI colors
@@ -276,7 +311,7 @@ Use `summarize --help` or `summarize help` for the full help text.
   - Install `uvx`: `brew install uv` (or https://astral.sh/uv/)
 - `--extract`: print extracted content and exit (URLs only; stdin `-` is not supported)
   - Deprecated alias: `--extract-only`
-- `--slides`: extract slides for YouTube/direct video URLs and render them inline in the summary narrative (auto-renders inline in supported terminals)
+- `--slides`: extract slides for YouTube, direct video URLs, or local video files and render them inline in the summary narrative (auto-renders inline in supported terminals)
 - `--slides-ocr`: run OCR on extracted slides (requires `tesseract`)
 - `--slides-dir <dir>`: base output dir for slide images (default `./slides`)
 - `--slides-scene-threshold <value>`: scene detection threshold (0.1-1.0)
@@ -286,7 +321,7 @@ Use `summarize --help` or `summarize help` for the full help text.
 - `--verbose`: debug/diagnostics on stderr
 - `--metrics off|on|detailed`: metrics output (default `on`)
 
-### Coding CLIs (Codex, Claude, Gemini, Agent)
+### Coding CLIs (Codex, Claude, Gemini, Agent, OpenClaw, OpenCode)
 
 Summarize can use common coding CLIs as local model backends:
 
@@ -294,11 +329,13 @@ Summarize can use common coding CLIs as local model backends:
 - `claude` -> `--cli claude` / `--model cli/claude/<model>`
 - `gemini` -> `--cli gemini` / `--model cli/gemini/<model>`
 - `agent` (Cursor Agent CLI) -> `--cli agent` / `--model cli/agent/<model>`
+- `openclaw` -> `--cli openclaw` / `--model cli/openclaw/<model>` or `--model openclaw/<model>`
+- `opencode` -> `--cli opencode` / `--model cli/opencode/<model>` (`--model cli/opencode` uses the OpenCode runtime default)
 
 Requirements:
 
-- Binary installed and on `PATH` (or set `CODEX_PATH`, `CLAUDE_PATH`, `GEMINI_PATH`, `AGENT_PATH`)
-- Provider authenticated (`codex login`, `claude auth`, `gemini` login flow, `agent login` or `CURSOR_API_KEY`)
+- Binary installed and on `PATH` (or set `CODEX_PATH`, `CLAUDE_PATH`, `GEMINI_PATH`, `AGENT_PATH`, `OPENCLAW_PATH`, `OPENCODE_PATH`)
+- Provider authenticated (`codex login`, `claude auth`, `gemini` login flow, `agent login` or `CURSOR_API_KEY`, `opencode auth login`)
 
 Quick smoke test:
 
@@ -309,13 +346,15 @@ summarize --cli codex --plain --timeout 2m /tmp/summarize-cli-smoke.txt
 summarize --cli claude --plain --timeout 2m /tmp/summarize-cli-smoke.txt
 summarize --cli gemini --plain --timeout 2m /tmp/summarize-cli-smoke.txt
 summarize --cli agent --plain --timeout 2m /tmp/summarize-cli-smoke.txt
+summarize --cli openclaw --plain --timeout 2m /tmp/summarize-cli-smoke.txt
+summarize --cli opencode --plain --timeout 2m /tmp/summarize-cli-smoke.txt
 ```
 
 Set explicit CLI allowlist/order:
 
 ```json
 {
-  "cli": { "enabled": ["codex", "claude", "gemini", "agent"] }
+  "cli": { "enabled": ["codex", "claude", "gemini", "agent", "openclaw", "opencode"] }
 }
 ```
 
@@ -327,7 +366,7 @@ Configure implicit auto CLI fallback:
     "autoFallback": {
       "enabled": true,
       "onlyWhenNoApiKeys": true,
-      "order": ["claude", "gemini", "codex", "agent"]
+      "order": ["claude", "gemini", "codex", "agent", "openclaw", "opencode"]
     }
   }
 }
@@ -343,7 +382,7 @@ CLI attempts are prepended when:
 - `cli.enabled` is set (explicit allowlist/order), or
 - implicit auto selection is active and `cli.autoFallback` is enabled.
 
-Default fallback behavior: only when no API keys are configured, order `claude, gemini, codex, agent`, and remember/prioritize last successful provider (`~/.summarize/cli-state.json`).
+Default fallback behavior: only when no API keys are configured, order `claude, gemini, codex, agent, openclaw, opencode`, and remember/prioritize last successful provider (`~/.summarize/cli-state.json`).
 
 Set explicit CLI attempts:
 
@@ -382,7 +421,8 @@ Non-YouTube URLs go through a fetch -> extract pipeline. When direct fetch/extra
 
 1. Apify (if `APIFY_API_TOKEN` is set): uses a scraping actor (`faVsWy9VTSNVIhWpR`)
 2. yt-dlp + Whisper (if `yt-dlp` is available): downloads audio, then transcribes with local `whisper.cpp` when installed
-   (preferred), otherwise falls back to OpenAI (`OPENAI_API_KEY`) or FAL (`FAL_KEY`)
+   (preferred), otherwise falls back to Groq (`GROQ_API_KEY`), AssemblyAI (`ASSEMBLYAI_API_KEY`), Gemini
+   (`GEMINI_API_KEY` / Google aliases), OpenAI (`OPENAI_API_KEY`), then FAL (`FAL_KEY`)
 
 Environment variables for yt-dlp mode:
 
@@ -390,19 +430,29 @@ Environment variables for yt-dlp mode:
 - `SUMMARIZE_WHISPER_CPP_MODEL_PATH` - optional override for the local `whisper.cpp` model file
 - `SUMMARIZE_WHISPER_CPP_BINARY` - optional override for the local binary (default: `whisper-cli`)
 - `SUMMARIZE_DISABLE_LOCAL_WHISPER_CPP=1` - disable local whisper.cpp (force remote)
+- `GROQ_API_KEY` - Groq Whisper transcription
+- `ASSEMBLYAI_API_KEY` - AssemblyAI transcription
+- `GEMINI_API_KEY` - Gemini transcription (`GOOGLE_GENERATIVE_AI_API_KEY` / `GOOGLE_API_KEY` also work)
 - `OPENAI_API_KEY` - OpenAI Whisper transcription
 - `OPENAI_WHISPER_BASE_URL` - optional OpenAI-compatible Whisper endpoint override
 - `FAL_KEY` - FAL AI Whisper fallback
 
 Apify costs money but tends to be more reliable when captions exist.
 
-### Slide extraction (YouTube + direct video URLs)
+### Slide extraction (YouTube + direct video URLs + local video files)
 
 Extract slide screenshots (scene detection via `ffmpeg`) and optional OCR:
 
+Requirements:
+
+- `ffmpeg` for scene detection and frame extraction
+- `yt-dlp` for YouTube video download/stream resolution
+- `tesseract` only when using `--slides-ocr`
+
 ```bash
 summarize "https://www.youtube.com/watch?v=..." --slides
 summarize "https://www.youtube.com/watch?v=..." --slides --slides-ocr
+summarize "/path/to/video.webm" --slides
 ```
 
 Outputs are written under `./slides/<sourceId>/` (or `--slides-dir`). OCR results are included in JSON output
@@ -411,7 +461,7 @@ extractor also samples at a fixed interval to improve coverage.
 When using `--slides`, supported terminals (kitty/iTerm/Konsole) render inline thumbnails automatically inside the
 summary narrative (the model inserts `[slide:N]` markers). Timestamp links are clickable when the terminal supports
 OSC-8 (YouTube/Vimeo/Loom/Dropbox). If inline images are unsupported, Summarize prints a note with the on-disk
-slide directory.
+slide directory. Local video files stay on the slide-aware path, transcribe in place, and avoid fake download labels.
 
 Use `--slides --extract` to print the full timed transcript and insert slide images inline at matching timestamps.
 
@@ -425,7 +475,7 @@ summarize "https://www.youtube.com/watch?v=..." --extract --format md --markdown
 
 Local audio/video files are transcribed first, then summarized. `--video-mode transcript` forces
 direct media URLs (and embedded media) through Whisper first. Prefers local `whisper.cpp` when available; otherwise requires
-`OPENAI_API_KEY` or `FAL_KEY`.
+one of `GROQ_API_KEY`, `ASSEMBLYAI_API_KEY`, `GEMINI_API_KEY` (or Google aliases), `OPENAI_API_KEY`, or `FAL_KEY`.
 
 ### Local ONNX transcription (Parakeet/Canary)
 
@@ -449,7 +499,7 @@ Run: `summarize <url>`
 - RSS feeds (Podcasting 2.0 transcripts when available)
 - Embedded YouTube podcast pages (e.g. JREPodcast)
 
-Transcription: prefers local `whisper.cpp` when installed; otherwise uses OpenAI Whisper or FAL when keys are set.
+Transcription: prefers local `whisper.cpp` when installed; otherwise uses Groq, AssemblyAI, Gemini, OpenAI, or FAL when keys are set.
 
 ### Translation paths
 
@@ -463,7 +513,7 @@ When the input is audio/video, the CLI needs a transcript first. The transcript
 2. Whisper transcription (fallback)
    - YouTube: falls back to yt-dlp (audio download) + Whisper transcription when configured; Apify is a last resort.
    - Prefers local `whisper.cpp` when installed + model available.
-   - Otherwise uses cloud Whisper (OpenAI `OPENAI_API_KEY`) or FAL (`FAL_KEY`).
+   - Otherwise uses cloud transcription in this order: Groq (`GROQ_API_KEY`) → AssemblyAI (`ASSEMBLYAI_API_KEY`) → Gemini (`GEMINI_API_KEY` / Google aliases) → OpenAI (`OPENAI_API_KEY`) → FAL (`FAL_KEY`).
 
 For direct media URLs, use `--video-mode transcript` to force transcribe -> summarize:
 
@@ -483,6 +533,7 @@ Supported keys today:
 {
   "model": { "id": "openai/gpt-5-mini" },
   "env": { "OPENAI_API_KEY": "sk-..." },
+  "output": { "length": "long" },
   "ui": { "theme": "ember" }
 }
 ```
@@ -502,6 +553,7 @@ Also supported:
 - `models` (define presets selectable via `--model <preset>`)
 - `env` (generic env var defaults; process env still wins)
 - `apiKeys` (legacy shortcut, mapped to env names; prefer `env` for new configs)
+- `output.length` (default `--length`: `short|medium|long|xl|xxl|20k`)
 - `cache.media` (media download cache: TTL 7 days, 2048 MB cap by default; `--no-media-cache` disables)
 - `media.videoMode: "auto"|"transcript"|"understand"`
 - `slides.enabled` / `slides.max` / `slides.ocr` / `slides.dir` (defaults for `--slides`)
@@ -552,6 +604,7 @@ Set the key matching your chosen `--model`:
   - legacy `"apiKeys"` still works (mapped to env names)
 
 - `OPENAI_API_KEY` (for `openai/...`)
+- `NVIDIA_API_KEY` (for `nvidia/...`)
 - `ANTHROPIC_API_KEY` (for `anthropic/...`)
 - `XAI_API_KEY` (for `xai/...`)
 - `Z_AI_API_KEY` (for `zai/...`; supports `ZAI_API_KEY` alias)
@@ -610,6 +663,7 @@ Example:
 
 ```bash
 OPENROUTER_API_KEY=sk-or-... summarize "https://example.com" --model openrouter/meta-llama/llama-3.1-8b-instruct:free
+OPENROUTER_API_KEY=sk-or-... summarize "https://example.com" --model openrouter/minimax/minimax-m2.5
 ```
 
 If your OpenRouter account enforces an allowed-provider list, make sure at least one provider
@@ -617,6 +671,18 @@ is allowed for the selected model. When routing fails, `summarize` prints the ex
 
 Legacy: `OPENAI_BASE_URL=https://openrouter.ai/api/v1` (and either `OPENAI_API_KEY` or `OPENROUTER_API_KEY`) also works.
 
+NVIDIA API Catalog (OpenAI-compatible; free credits):
+
+- Set `NVIDIA_API_KEY=...`
+- Optional: `NVIDIA_BASE_URL=https://integrate.api.nvidia.com/v1`
+- Credits: API Catalog trial starts with 1000 free API credits on signup (up to 5000 total via “Request More” in the API Catalog profile)
+- Pick a model id from `/v1/models` (examples: fast `stepfun-ai/step-3.5-flash`, strong but slower `z-ai/glm5`)
+
+```bash
+export NVIDIA_API_KEY="nvapi-..."
+summarize "https://example.com" --model nvidia/stepfun-ai/step-3.5-flash
+```
+
 Z.AI (OpenAI-compatible):
 
 - `Z_AI_API_KEY=...` (or `ZAI_API_KEY=...`)
@@ -626,6 +692,10 @@ Optional services:
 
 - `FIRECRAWL_API_KEY` (website extraction fallback)
 - `YT_DLP_PATH` (path to yt-dlp binary for audio extraction)
+- `GROQ_API_KEY` (Groq Whisper transcription)
+- `ASSEMBLYAI_API_KEY` (AssemblyAI transcription)
+- `GEMINI_API_KEY` / `GOOGLE_GENERATIVE_AI_API_KEY` / `GOOGLE_API_KEY` (Gemini transcription)
+- `OPENAI_API_KEY` / `OPENAI_WHISPER_BASE_URL` (OpenAI Whisper transcription)
 - `FAL_KEY` (FAL AI API key for audio transcription via Whisper)
 - `APIFY_API_TOKEN` (YouTube transcript fallback)
 

package/dist/cli.js

@@ -1,3 +1,3 @@
 #!/usr/bin/env node
-if (!process.env.SUMMARIZE_GIT_SHA) process.env.SUMMARIZE_GIT_SHA = "ae52818b"
+if (!process.env.SUMMARIZE_GIT_SHA) process.env.SUMMARIZE_GIT_SHA = "70006c3e"
 await import('./esm/cli.js')

package/dist/esm/cache-keys.js

@@ -0,0 +1,83 @@
+import { createHash } from "node:crypto";
+export function hashString(value) {
+    return createHash("sha256").update(value).digest("hex");
+}
+export function hashJson(value) {
+    return hashString(JSON.stringify(value));
+}
+export function normalizeContentForHash(content) {
+    return content.replaceAll("\r\n", "\n").trim();
+}
+export function extractTaggedBlock(prompt, tag) {
+    const open = `<${tag}>`;
+    const close = `</${tag}>`;
+    const start = prompt.indexOf(open);
+    if (start === -1)
+        return null;
+    const end = prompt.indexOf(close, start + open.length);
+    if (end === -1)
+        return null;
+    return prompt.slice(start + open.length, end).trim();
+}
+export function buildPromptHash(prompt) {
+    const instructionsContent = extractTaggedBlock(prompt, "instructions");
+    const contextContent = extractTaggedBlock(prompt, "context");
+    // If at least one of the tags is present (even if empty), hash their contents.
+    if (instructionsContent !== null || contextContent !== null) {
+        const instructions = instructionsContent ?? "";
+        const context = contextContent ?? "";
+        return hashString(`${instructions}\n${context}`.trim());
+    }
+    // Fallback for prompts without any tags.
+    return hashString(prompt.trim());
+}
+export function buildPromptContentHash({ prompt, fallbackContent, }) {
+    const content = extractTaggedBlock(prompt, "content") ?? fallbackContent ?? null;
+    if (!content || content.trim().length === 0)
+        return null;
+    return hashString(normalizeContentForHash(content));
+}
+export function buildLengthKey(lengthArg) {
+    return lengthArg.kind === "preset"
+        ? `preset:${lengthArg.preset}`
+        : `chars:${lengthArg.maxCharacters}`;
+}
+export function buildLanguageKey(outputLanguage) {
+    return outputLanguage.kind === "auto" ? "auto" : outputLanguage.tag;
+}
+export function buildExtractCacheKeyValue({ url, options, formatVersion, }) {
+    return hashJson({ url, options, formatVersion });
+}
+export function buildSummaryCacheKeyValue({ contentHash, promptHash, model, lengthKey, languageKey, formatVersion, }) {
+    return hashJson({
+        contentHash,
+        promptHash,
+        model,
+        lengthKey,
+        languageKey,
+        formatVersion,
+    });
+}
+export function buildSlidesCacheKeyValue({ url, settings, formatVersion, }) {
+    return hashJson({
+        url,
+        settings: {
+            ocr: settings.ocr,
+            outputDir: settings.outputDir,
+            sceneThreshold: settings.sceneThreshold,
+            autoTuneThreshold: settings.autoTuneThreshold,
+            maxSlides: settings.maxSlides,
+            minDurationSeconds: settings.minDurationSeconds,
+        },
+        formatVersion,
+    });
+}
+export function buildTranscriptCacheKeyValue({ url, namespace, formatVersion, fileMtime, }) {
+    return hashJson({
+        url,
+        namespace,
+        fileMtime: fileMtime ?? null,
+        formatVersion,
+    });
+}
+//# sourceMappingURL=cache-keys.js.map

package/dist/esm/cache-keys.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"cache-keys.js","sourceRoot":"","sources":["../../src/cache-keys.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,UAAU,EAAE,MAAM,aAAa,CAAC;AAIzC,MAAM,UAAU,UAAU,CAAC,KAAa;IACtC,OAAO,UAAU,CAAC,QAAQ,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;AAC1D,CAAC;AAED,MAAM,UAAU,QAAQ,CAAC,KAAc;IACrC,OAAO,UAAU,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,CAAC;AAC3C,CAAC;AAED,MAAM,UAAU,uBAAuB,CAAC,OAAe;IACrD,OAAO,OAAO,CAAC,UAAU,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC,IAAI,EAAE,CAAC;AACjD,CAAC;AAED,MAAM,UAAU,kBAAkB,CAChC,MAAc,EACd,GAA2C;IAE3C,MAAM,IAAI,GAAG,IAAI,GAAG,GAAG,CAAC;IACxB,MAAM,KAAK,GAAG,KAAK,GAAG,GAAG,CAAC;IAC1B,MAAM,KAAK,GAAG,MAAM,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;IACnC,IAAI,KAAK,KAAK,CAAC,CAAC;QAAE,OAAO,IAAI,CAAC;IAC9B,MAAM,GAAG,GAAG,MAAM,CAAC,OAAO,CAAC,KAAK,EAAE,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,CAAC;IACvD,IAAI,GAAG,KAAK,CAAC,CAAC;QAAE,OAAO,IAAI,CAAC;IAC5B,OAAO,MAAM,CAAC,KAAK,CAAC,KAAK,GAAG,IAAI,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;AACvD,CAAC;AAED,MAAM,UAAU,eAAe,CAAC,MAAc;IAC5C,MAAM,mBAAmB,GAAG,kBAAkB,CAAC,MAAM,EAAE,cAAc,CAAC,CAAC;IACvE,MAAM,cAAc,GAAG,kBAAkB,CAAC,MAAM,EAAE,SAAS,CAAC,CAAC;IAE7D,+EAA+E;IAC/E,IAAI,mBAAmB,KAAK,IAAI,IAAI,cAAc,KAAK,IAAI,EAAE,CAAC;QAC5D,MAAM,YAAY,GAAG,mBAAmB,IAAI,EAAE,CAAC;QAC/C,MAAM,OAAO,GAAG,cAAc,IAAI,EAAE,CAAC;QACrC,OAAO,UAAU,CAAC,GAAG,YAAY,KAAK,OAAO,EAAE,CAAC,IAAI,EAAE,CAAC,CAAC;IAC1D,CAAC;IAED,yCAAyC;IACzC,OAAO,UAAU,CAAC,MAAM,CAAC,IAAI,EAAE,CAAC,CAAC;AACnC,CAAC;AAED,MAAM,UAAU,sBAAsB,CAAC,EACrC,MAAM,EACN,eAAe,GAIhB;IACC,MAAM,OAAO,GAAG,kBAAkB,CAAC,MAAM,EAAE,SAAS,CAAC,IAAI,eAAe,IAAI,IAAI,CAAC;IACjF,IAAI,CAAC,OAAO,IAAI,OAAO,CAAC,IAAI,EAAE,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,IAAI,CAAC;IACzD,OAAO,UAAU,CAAC,uBAAuB,CAAC,OAAO,CAAC,CAAC,CAAC;AACtD,CAAC;AAED,MAAM,UAAU,cAAc,CAAC,SAAoB;IACjD,OAAO,SAAS,CAAC,IAAI,KAAK,QAAQ;QAChC,CAAC,CAAC,UAAU,SAAS,CAAC,MAAM,EAAE;QAC9B,CAAC,CAAC,SAAS,SAAS,CAAC,aAAa,EAAE,CAAC;AACzC,CAAC;AAED,MAAM,UAAU,gBAAgB,CAAC,cAA8B;IAC7D,OAAO,cAAc,CAAC,IAAI,KAAK,MAAM,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,cAAc,CAAC,GAAG,CAAC;AACtE,CAAC;AAED,MAAM,UAAU,yBAAyB,CAAC,EACxC,GAAG,EACH,OAAO,EACP,aAAa,GAKd;IACC,OAAO,QAAQ,CAAC,EAAE,GAAG,EAAE,OAAO,EAAE,aAAa,EAAE,CAAC,CAAC;AACnD,CAAC;AAED,MAAM,UAAU,yBAAyB,CAAC,EACxC,WAAW,EACX,UAAU,EACV,KAAK,EACL,SAAS,EACT,WAAW,EACX,aAAa,GAQd;IACC,OAAO,QAAQ,CAAC;QACd,WAAW;QACX,UAAU;QACV,KAAK;QACL,SAAS;QACT,WAAW;QACX,aAAa;KACd,CAAC,CAAC;AACL,CAAC;AAED,MAAM,UAAU,wBAAwB,CAAC,EACvC,GAAG,EACH,QAAQ,EACR,aAAa,GAYd;IACC,OAAO,QAAQ,CAAC;QACd,GAAG;QACH,QAAQ,EAAE;YACR,GAAG,EAAE,QAAQ,CAAC,GAAG;YACjB,SAAS,EAAE,QAAQ,CAAC,SAAS;YAC7B,cAAc,EAAE,QAAQ,CAAC,cAAc;YACvC,iBAAiB,EAAE,QAAQ,CAAC,iBAAiB;YAC7C,SAAS,EAAE,QAAQ,CAAC,SAAS;YAC7B,kBAAkB,EAAE,QAAQ,CAAC,kBAAkB;SAChD;QACD,aAAa;KACd,CAAC,CAAC;AACL,CAAC;AAED,MAAM,UAAU,4BAA4B,CAAC,EAC3C,GAAG,EACH,SAAS,EACT,aAAa,EACb,SAAS,GAMV;IACC,OAAO,QAAQ,CAAC;QACd,GAAG;QACH,SAAS;QACT,SAAS,EAAE,SAAS,IAAI,IAAI;QAC5B,aAAa;KACd,CAAC,CAAC;AACL,CAAC"}

package/dist/esm/cache-slides-cleanup.js

@@ -0,0 +1,47 @@
+import { rmSync } from "node:fs";
+import { isAbsolute, join, resolve as resolvePath, sep as pathSep } from "node:path";
+function normalizeAbsolutePath(value) {
+    if (typeof value !== "string")
+        return null;
+    const trimmed = value.trim();
+    if (!trimmed)
+        return null;
+    const resolved = resolvePath(trimmed);
+    return isAbsolute(resolved) ? resolved : null;
+}
+export function cleanupSlidesPayload(raw) {
+    let payload;
+    try {
+        payload = JSON.parse(raw);
+    }
+    catch {
+        return;
+    }
+    if (!payload || typeof payload !== "object")
+        return;
+    const slidesDir = normalizeAbsolutePath(payload.slidesDir);
+    const slides = Array.isArray(payload.slides)
+        ? payload.slides
+        : [];
+    if (!slidesDir)
+        return;
+    const dirPrefix = slidesDir.endsWith(pathSep) ? slidesDir : `${slidesDir}${pathSep}`;
+    const safeRemove = (target) => {
+        try {
+            rmSync(target, { force: true });
+        }
+        catch {
+            // ignore
+        }
+    };
+    for (const slide of slides) {
+        const imagePath = normalizeAbsolutePath(slide?.imagePath);
+        if (!imagePath)
+            continue;
+        if (!imagePath.startsWith(dirPrefix))
+            continue;
+        safeRemove(imagePath);
+    }
+    safeRemove(join(slidesDir, "slides.json"));
+}
+//# sourceMappingURL=cache-slides-cleanup.js.map

package/dist/esm/cache-slides-cleanup.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"cache-slides-cleanup.js","sourceRoot":"","sources":["../../src/cache-slides-cleanup.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,MAAM,EAAE,MAAM,SAAS,CAAC;AACjC,OAAO,EAAE,UAAU,EAAE,IAAI,EAAE,OAAO,IAAI,WAAW,EAAE,GAAG,IAAI,OAAO,EAAE,MAAM,WAAW,CAAC;AAErF,SAAS,qBAAqB,CAAC,KAAc;IAC3C,IAAI,OAAO,KAAK,KAAK,QAAQ;QAAE,OAAO,IAAI,CAAC;IAC3C,MAAM,OAAO,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC;IAC7B,IAAI,CAAC,OAAO;QAAE,OAAO,IAAI,CAAC;IAC1B,MAAM,QAAQ,GAAG,WAAW,CAAC,OAAO,CAAC,CAAC;IACtC,OAAO,UAAU,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC;AAChD,CAAC;AAED,MAAM,UAAU,oBAAoB,CAAC,GAAW;IAC9C,IAAI,OAAgB,CAAC;IACrB,IAAI,CAAC;QACH,OAAO,GAAG,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;IAC5B,CAAC;IAAC,MAAM,CAAC;QACP,OAAO;IACT,CAAC;IACD,IAAI,CAAC,OAAO,IAAI,OAAO,OAAO,KAAK,QAAQ;QAAE,OAAO;IACpD,MAAM,SAAS,GAAG,qBAAqB,CAAE,OAAmC,CAAC,SAAS,CAAC,CAAC;IACxF,MAAM,MAAM,GAAG,KAAK,CAAC,OAAO,CAAE,OAAgC,CAAC,MAAM,CAAC;QACpE,CAAC,CAAG,OAAgC,CAAC,MAAyC;QAC9E,CAAC,CAAC,EAAE,CAAC;IACP,IAAI,CAAC,SAAS;QAAE,OAAO;IACvB,MAAM,SAAS,GAAG,SAAS,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,GAAG,SAAS,GAAG,OAAO,EAAE,CAAC;IACrF,MAAM,UAAU,GAAG,CAAC,MAAc,EAAE,EAAE;QACpC,IAAI,CAAC;YACH,MAAM,CAAC,MAAM,EAAE,EAAE,KAAK,EAAE,IAAI,EAAE,CAAC,CAAC;QAClC,CAAC;QAAC,MAAM,CAAC;YACP,SAAS;QACX,CAAC;IACH,CAAC,CAAC;IACF,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;QAC3B,MAAM,SAAS,GAAG,qBAAqB,CAAC,KAAK,EAAE,SAAS,CAAC,CAAC;QAC1D,IAAI,CAAC,SAAS;YAAE,SAAS;QACzB,IAAI,CAAC,SAAS,CAAC,UAAU,CAAC,SAAS,CAAC;YAAE,SAAS;QAC/C,UAAU,CAAC,SAAS,CAAC,CAAC;IACxB,CAAC;IACD,UAAU,CAAC,IAAI,CAAC,SAAS,EAAE,aAAa,CAAC,CAAC,CAAC;AAC7C,CAAC"}