📥 DSH Chat Import

Import 15+ external agent conversation histories into DeepSeek Harness as full-fidelity, resumable sessions — and export / sync back to Claude Code, Codex, Kimi, or a portable interchange bundle.

💡 Intro · 🚀 Install · ✨ Features · 🗂 Supported sources · 🛠 Usage · 🔑 Key behaviors · 📚 Docs · ⭐ Star History · 🤝 Contributing

16+ agent sources, one plugin — full-fidelity import into DeepSeek Harness, seamless resume, matrix interop / backup / handoff on the way out.

Changelog: CHANGELOG.md · Roadmap: ROADMAP.md · Interchange protocol: docs/INTERCHANGE.md

💡 Intro

dsh-chat-import imports conversation histories from Claude Code, Codex, ChatGPT, Cursor, Gemini, Reasonix, opencode, MiMo Code, ZCode, Grok Build, OpenClaw, Pi Coding Agent, Hermes, Kimi CLI / Kimi Code, Qoder CLI and DSH session logs — tool calls, reasoning and all — as full-fidelity, resumable DeepSeek Harness sessions. Source files are read read-only (never rewritten), the DSH engine is never touched, and every import becomes a fresh session grouped into the workspace of its source cwd.

The reverse direction is covered too: export_claude serializes a DSH session back into a Claude Code JSONL transcript that Claude Code can load with --resume (read-only — your DSH log is never modified), sync_to_claude incrementally appends a session's new turns back to a Claude Code file — guarded, never silently overwriting — and the same matrix extends to Codex rollouts (export_codex) and Kimi wire files (export_kimi), plus a portable interchange bundle (export_bundle / restore_bundle) with SHA-256 fingerprints and cross-machine restore.

Requires Node.js ≥ 22.13, targets dsh 0.1.x (tested on 0.1.0-rc.6 / 0.1.0-rc.7).

🚀 Install

dsh plugin --profile web add dsh-chat-import                    # npm package
dsh plugin --profile web add -w link:/path/to/dsh-chat-import   # local checkout (symlink)

Then:

1. Import — call any import_* tool in a DSH session:

import_claude({ path: "~/.claude/projects" })
import_chatgpt({ path: "~/Downloads/chatgpt-export/conversations.json" })
import_local_jsonl({ path: "D:\downloads\session.jsonl" })

2. Resume — refresh the session list, open the imported session, and keep chatting from where the source left off.
3. Discover & batch — scan_discover() previews read-only; the sidebar "Import sessions" panel browses by workspace and supports multi-select import; /import-all batch-imports everything.
4. Sync (optional) — the panel's "Sync" tab offers bidirectional incremental sync (external → DSH, DSH → external), default off. Sub-agent conversations are filtered out by default in both directions, and excludeDirs deny-lists let you skip specific workspace directories per direction.

Uninstall: remove the import-claude insert line from your profile's bundles and restart dsh; imported sessions stay untouched and the plugin never auto-deletes.

✨ Features

🗂 Supported sources

Each import preserves what the source actually records; anything a format cannot preserve is explicitly flagged in the import report. Per-format details and edge cases live in Usage Reference.

🛠 Usage

All import_* tools share the same path semantics: a single file becomes one session, a directory is scanned recursively for batch import. Common options: preview (zero side effects), force (new full copy), sessionId (override target id), expectedHash (SHA-256 verification), restamp (shift timestamps to now), workspaceMode / workspaceDir (grouping control).

import_claude({ path: "C:\Users\<you>\.claude\projects\<slug>\<sessionId>.jsonl" })
import_opencode({ path: "C:\Users\<you>\.local\share\opencode\opencode.db" })
import_local_jsonl({ path: "D:\downloads\session.jsonl", format: "claude" })

import_chatgpt / import_opencode / import_zcode / import_hermes always return batch results — one file / database contains every session, and each conversation becomes its own session in a single call.

Full per-tool / per-command usage lives in docs/USAGE.md.

🔑 Key behaviors

• Read-only import — source transcripts and databases are never rewritten; imported DSH history is append-only.
• Idempotent + incremental — unchanged sources skip without re-reading; grown sources append only new turns; shrinking is detected and reported.
• Auto workspace grouping — sessions land in the workspace of their source cwd (authoritative mapping → slug decode → home-directory sandbox guard; falls back to the source file's directory when the path does not exist locally).
• Preset mode — imported sessions mount the default preset scope via agents.create and write the default preset id back to SessionHeader.agentPreset, so the UI shows the preset-mode chip exactly like a normal session.
• System prompt (optional, off by default) — the "Import system prompt" setting (a tab in the settings Plugins section) preserves the source transcript's system / developer prompt as a "context injection" collapsed row, prefixed with a note that the environment changed and tools / permissions / instructions now follow DSH. Claude Code transcripts do not persist a system prompt, so the toggle is a no-op for that source.
• Fail loudly — malformed lines, suspected secrets, format limitations and export degradations are all reported; every persisted session gets a structural self-check.
• Sandbox — reading sources or writing exports outside the workspace requires the session sandbox to allow that path.

📚 Docs

⭐ Star History

🤝 Contributing

Contributions welcome — fork the repo, create a feature/<name> branch, and open a PR. Full guide: CONTRIBUTING.md.

• Tests: npm test · Cross-platform guard: npm run check:linux
• Repo conventions: AGENTS.md — conventional commits, bilingual READMEs must stay in sync, plugin consumes public dsh host services only, multi-session file-claim protocol.

📄 License

MIT — see LICENSE.