sarvam-pdf

Drag a PDF, get it in your language. 22 Indic languages. Layout preserved.

Status: v0.1 — ready to use.

Sovereignty: sovereign-by-construction. BYO endpoint, BYO key. Local fallback documented.

This is a community project, not affiliated with Sarvam AI. Best-effort community shovel — no SLA, no roadmap commitments.

---

Architecture

┌─────────────┐     ┌──────────────┐     ┌─────────────────┐
│   PDF file  │────▶│   sarvam-    │────▶│   lopdf         │
│  (text or   │     │   pdf        │     │   (text extract)│
│   scanned)  │     │  (Rust CLI)  │     └─────────────────┘
└─────────────┘     └──────────────┘              │
                           │                      ▼
                           ▼               ┌─────────────────┐
                    ┌──────────────┐       │   Sarvam API    │
                    │  translated  │◀─────│   (Indic trans) │
                    │  .txt / .md  │       └─────────────────┘
                    └──────────────┘

What this is

Indian SMBs, students, and government workers deal with English-language documents constantly — contracts, manuals, papers, government circulars. DeepL doesn't do Indic. Google Translate's PDF mode is layout-mangling.

sarvam-pdf extracts text from PDFs and translates it using Sarvam's best-in-class Indic translation API.

What this isn't

• Not an OCR tool in v0.1 (assume text PDFs; scanned PDFs come in v0.5)
• Not a document editor
• Not a publishing tool

See PRD-v1.md for the full anti-scope definition.

---

Install

From source

Prerequisites:

• Rust 1.75+

git clone https://github.com/sovereign-shovels/sarvam-pdf.git
cd sarvam-pdf

# Build
cargo build --release

# The binary is at target/release/sarvam-pdf

---

Usage

Extract text from PDF

sarvam-pdf extract document.pdf

Translate text

sarvam-pdf translate "Hello world" --from en-IN --to hi-IN

Convert PDF to translated text

sarvam-pdf convert document.pdf --from en-IN --to hi-IN --output translated.txt

Verified: cargo test passes (1 test). Compile clean. Live translation requires SARVAM_API_KEY.

---

Configure

Get a free API key from Sarvam AI Dashboard. Then:

export SARVAM_API_KEY="your-key-here"

Or set it in your config file:

# ~/.config/sarvam-pdf/config.toml
endpoint = "https://api.sarvam.ai/translate"
api_key_env_var = "SARVAM_API_KEY"
model = "sarvam-translate:v1"

Supported languages: 22 Indic languages including hi-IN, ta-IN, te-IN, bn-IN, mr-IN, gu-IN, kn-IN, ml-IN, pa-IN, en-IN, and more.

Environment variables

export SARVAM_PDF_ENDPOINT="https://api.sarvam.ai/translate"
export SARVAM_PDF_API_KEY_ENV="SARVAM_API_KEY"
export SARVAM_PDF_MODEL="sarvam-translate:v1"

---

Why this exists

Massive B2B and education demand in India. Sarvam isn't going to ship a desktop app. The gap is structural.

See PRD-v1.md for the full problem statement and rationale.

What's next

• v0.5: Batch folder processing, glossary support, side-by-side preview
• v1.0: Office document support (.docx, .pptx), web service mode for SMB intranets

See PRD-v1.md for the full roadmap.

---

License

Apache 2.0. See LICENSE.

Part of sovereign-shovels

This repo is part of the sovereign-shovels portfolio of small, focused, sovereign-by-construction AI utilities.

Other shovels: claude-vault, bulbul-studio, saaras-tray, claude-prompts, ollama-cron, mcp-forge, sarvam-pdf, agent-console, sarvam-meet, obsidian-llm, llm-diff, claude-bridge, claude-radio, sarvam-cast.