Tocify

Tocify is a free, online and open-source web application to add, edit, or generate Table of Contents (ToCs) / bookmarks for PDFs. It uses AI to parse scanned directory images or raw text into structured outlines, and creates clickable bookmarks & printable TOC pages for PDFs in cross platforms.

AI features require your own provider API key. Enter it in API Settings; it is stored in your browser and requests go directly to your selected provider. PDF editing and local OCR remain free.

Local development and model testing

Run pnpm install and pnpm dev, then open http://localhost:5173. Load a PDF, choose a provider in API Settings, and enter your new key. There is no Auto provider or shared server key.

Default models are Gemini gemini-3.5-flash-lite, Qwen qwen3.8-flash, Zhipu text glm-4.7-flash and vision glm-4.6v-flash, DeepSeek deepseek-flash (text and images), and custom OpenAI gpt-6-luna. Model fields suggest newer alternatives and accept custom model IDs. Doubao defaults to doubao-seed-2-1-lite-260915 for text and images. Activate this model in Ark first; the model fields also accept your existing Endpoint IDs.

pnpm test:llm verifies missing-key rejection and model routing with mocked responses, without API usage. For real API tests, copy .env.example to .env.local and fill in the desired TOCIFY_TEST_* keys. These variables are only read by the test script and are never used to fund website requests.

pnpm test:llm:live --provider gemini
pnpm test:llm:live --provider deepseek --image /absolute/path/to/toc.png
pnpm test:llm:live --provider zhipu --image /absolute/path/to/toc.png

Live tests call the provider and use its API quota. Omitting --provider tests all configured test keys. Use --model MODEL_ID with --provider to test another model. Text tests check chapter names, hierarchy, and page numbers; image tests check that the supplied image produces a nonempty outline.

The built-in Qwen provider uses the Beijing API endpoint. Use a Beijing-region key; for another region, select Custom / OpenAI Compatible and enter that region's base URL. A 403 Access denied by API-Key restrictions response calls for checking the key's model permissions and IP allowlist in Model Studio. Changing the model ID alone does not grant access to the new model. Use the exact lowercase ID, for example qwen3.8-flash. See API key permissions.

Vercel configuration

The website does not require model credentials or Redis rate-limit credentials on Vercel. Deploy the updated code and confirm that /api/process-toc and /api/generate-board return HTTP 400 when no user API key is supplied. You can then remove these obsolete environment variables from Production, Preview, and Development:

AI_PROVIDER, GOOGLE_API_KEY, DASHSCOPE_API_KEY, QWEN_VL_MODEL, ZHIPU_API_KEY, DOUBAO_API_KEY, DOUBAO_ENDPOINT_ID_TEXT, DOUBAO_ENDPOINT_ID_VISION, UPSTASH_REDIS_REST_URL, and UPSTASH_REDIS_REST_TOKEN.

Vercel environment-variable changes apply to new deployments. Previous deployments keep their existing configuration; retire obsolete deployments and revoke former shared provider keys after migrating. Keep new test keys in your local .env.local or browser API Settings.