Open Source · Tauri 2 · 25 Tools · MCP · Roleplay · Voice Cloning · Cache-First · Code Review

CodePapr — Vibe Coding, Cache-First

CodePapr is a DeepSeek cache-friendly local coding Agent system. It's not just a code generator — it's an intelligent agent workbench with understanding + execution capabilities: reading projects, searching code, running commands, modifying files, previewing results, all locally. Also supports character roleplay and local voice cloning (GPT-SoVITS) so your Agent speaks in your favorite character's style. DeepSeek prefix caching maximizes token savings.

codepapr — Type "/" to open command palette
/review src/auth/login.ts
/reviewReview current changes or specified scope
/fixLocate and fix issues, auto-verify
/newCreate new file, component, or feature from description
/searchSearch codebase (fast model)
/commitStage changes and generate commit message (fast model)
/lintRun linter and fix violations (fast model)
→ Reviewing src/auth/login.ts ...
✓ No blocking issues found
Suggestion: add exception-handling test for password reset flow

One Runtime, Three Modes

From understanding to execution, switch modes without losing context

Ask

Explain architecture, trace responsibilities, analyze error causes. Read-only by default — no file changes or commands.

Read-Only · Explain

Plan

Before big changes, produce task lists, affected files, and verification paths. Break complex tasks into clear steps before acting.

Plan · Decompose

Agent

Fully autonomous: search project, read/write files, run commands, verify. Fix bugs and implement features in one go.

Execute · Read/Write

More Than a Chat Window

File Tree · Monaco Editor · Git · Browser · LSP — all in one interface

Project File Tree + LSP

Full project browsing, Monaco syntax highlighting, real LSP hover, jump-to-definition, and diagnostics.

Slash Command Palette

Type / to open command list, keyboard navigation + hover details, zero learning curve.

Agent Autonomous Execution

Auto search + read/write + run commands + verify. Runs in Web Worker, UI stays responsive.

Deep Git Integration

Diff, commit, branch switch, safe rollback. Agent handles analysis; human gates high-risk operations.

Browser Integration

Launch preview, page interactions, screenshots. Frontend debugging in one window.

TodoList Task Planning

Auto-create checklist for complex tasks, track progress in real-time, auto-collapse on completion. Auto-retry on failure.

Conversation Reset (Git)

Auto git commit snapshot before each user message. Hover any message to one-click reset --hard rollback, code and conversation sync recovery.

Auto Project Memory

Accumulates user profile, error patterns, project conventions across sessions. Async dedup & merge with fast model when >200 lines; non-blocking, uses consolidated version on next launch.

Visual Code Review

Built-in diff review panel, per-line comments, approval status, unresolved count. Human review after agent changes at a glance.

Character Roleplay

Create, import (CCv3 character card PNG/JSON), export AI characters. Agent speaks in character persona when activated, with full support for name, avatar, personality, scenario, opening message, example dialogue, and more.

Local Voice Cloning (TTS)

GPT-SoVITS powered voice cloning, replicate character voices with just 3-10 seconds of reference audio. Current UI uses WebSocket batch streaming by default, ~1-2s first-token latency; supports speed, synthesis rate (4/8/16), sentence merging, and voice fine-tuning.

Built from Scratch by CodePapr

All projects below were independently designed, coded, and organized by CodePapr

Type / to Invoke

18 built-in commands, auto-routed to primary or fast model by task complexity; /help /commands /compact bypass models, zero token cost

CommandPurposeModel
/help   /commandsList all available commands and descriptionsLocal
/compactForce-compress current session context, generate checkpointLocal
/reviewReview changes or specified scope, prioritize bugs and regression risksPrimary
/fixLocate and fix issues, auto-run verificationPrimary
/testAdd tests for specified feature or run existing testsPrimary
/explainExplain file, function, symbol, or error implementation approachPrimary
/diagnoseDiagnose errors, slow operations, or abnormal behavior root causesPrimary
/refactorSafe refactoring, preserving behaviorPrimary
/docWrite or update documentation (.md extension only)Primary
/newCreate new file, component, or feature from descriptionPrimary
/optimizeAnalyze and fix performance bottlenecks, reduce complexity or resource usagePrimary
/searchSearch codebase for patterns, usage, definitions, or referencesFast
/lintRun linter and fix violationsFast
/cleanClean dead code, unused imports, and leftover debug statementsFast
/commitStage changes and generate standardized commit messageFast
/summaryHigh-level overview of file, module, or entire projectFast
/buildBuild project and diagnose/fix build errorsFast

Desktop App + Cloud Intelligence

Tauri 2 Desktop Shell + React Workbench + Rust Backend + DeepSeek / OpenAI / Claude

Tauri 2

macOS · Windows
Native Window & Process

Rust Backend

Workspace IO · LSP
Async Commands · Search

Core Agent

Cache Partitioning · Sessions
25 Tools · 3 Sub-Agents
MCP · TodoList · Reset

React UI

Monaco · Git Panel
Browser · Light/Dark

Ready to Give Your Code an Agent?

Open source, runs locally, zero configuration.