dsh-continuum
Persistent project memory, context-safe agent handoffs, checkpointing, evidence tracking, and long-running task continuity for DeepSeek Harness
407 results
Persistent project memory, context-safe agent handoffs, checkpointing, evidence tracking, and long-running task continuity for DeepSeek Harness
Bridges images into text for DeepSeek Harness: when the session's selected model cannot see images, a configured vision model describes them and the descriptions enter the durable session history as folded context rows — your message stays untouched.
OpenViking memory & context bundle for DeepSeek Harness: auto-recall, session capture, threshold commit, offline outbox replay, self-contained MCP tool bridge, viking:// URI guard.
Tiered micro-clear layer for DeepSeek Harness compaction: clears whitelisted re-runnable tool results to placeholders before the official summarization tier, freeing context with no model call.
DeepSeek Harness (dsh) bundle that routes the harness through the OpenLLM gateway — an `openllm` provider on the in-box pi-ai adapter (via the local daemon gateway) plus the OpenLLM MCP server. Pure config; install OpenLLM first (see README).
Keeps long coding-agent sessions lean: wrap work in named tasks and, when one is done, fold its whole span into a short titled summary. The conversation stays readable, context costs stay low, and every fold's original content can be read back on demand. For DeepSeek Harness.
Out-of-tree dsh plugin: GitLab, GitHub, Gitee, Gitea, and Bitbucket tokens stay out of the model context, stored encrypted (AES-256-GCM) in a plugin-owned file; the model calls forge API tools on demand, and the web settings page manages sites and tokens.
Conversation-local credential handoff that writes through the DSH credential service without returning the secret to model context.
Curate tool output before it enters the model's context: keep error/warning lines and head/tail, spill the full text to disk with a pointer. Faster turns, smaller context, sharper attention — for every tool, not just bash.
DSH web plugin: Codex-style /side side conversations. Open a temporary fork of the current chat in a floating panel — ask side questions without interrupting the main task or polluting its context.
最新文档注入防幻觉:编写/引用库代码前先拉取版本化最新文档(ctx7 CLI 或 MCP),library ID 直取、版本指定、规则自动触发;消灭过期示例与幻觉 API。受 upstash/context7(61k★ MIT)启发。
DeepSeek Harness bundle that registers the session-summarize skill pack on ctx.skills.
A DeepSeek Harness plugin that integrates claude-mem: query, context injection, manual save, and session lifecycle over a local claude-mem worker's HTTP API
Vision for text-only DeepSeek Harness agents: delegate image reading to a one-shot subagent on a configurable vision route (MiniMax / Kimi / any OpenAI-compatible provider), keeping image bytes and the vision model's context out of the main session.
MCP 服务器管理器:在 DSH 设置页配置 MCP 服务器,宿主侧按配置热挂载/卸载 mcp-client 实例
Agentic Surface Compaction (ASC) for DeepSeek Harness: the model decides when and what to compact, committed as durable session-log replacements with full replay, search, and degradation
DeepSeek Harness plugin: cross-session memory for the model. Full-text search all past sessions (SQLite FTS5 via ctx.sessionQuery) and bring the strongest matching excerpts back into the current context — recall earlier decisions, commands, and mistakes.
上下文治理一体化插件:/context 命令 + ctx.contextMeter 服务 + 剪枝工具(prune_candidates/apply/expand/guard/stats)+ 两条提醒通道(上下文提醒 / 压缩告警,状态落盘跨重启去重)
DSH 会话上下文统计面板:统计每个会话的消息数、轮次、工具调用、字符数与预估 token,生成全量/按项目上下文报告,帮你看清上下文消耗大户。Context usage dashboard for DeepSeek Harness — per-session message/token statistics and context reports (zstd + plaintext logs).
Focus board for DeepSeek Harness agents: durable, model-maintained notes in the session workspace that pin the objective, constraints, and decisions across compaction and sessions — with structured entry tags (kind/status), a grouped summary view, automatic context injection, an archive on clear, and an optional read-only web panel
Cairn — long-term memory for the DeepSeek Harness: persistent cross-session memory (store, tools, auto-extraction, context injection) as one installable profile bundle
MCP 服务器可视化配置插件:在 DSH Web 设置界面管理 MCP 服务器(stdio / streamable-http / SSE),保存即生效,无需手改配置文件。| Visual MCP server management for DeepSeek Harness: configure stdio / HTTP / SSE servers from the settings UI, hot-synced without editing config files.
Real-time context occupancy meter for DSH Web: draggable compact/detail panel, per-tool usage, settings-editable cost table (CNY/USD, flat or peak/off-peak), per-provider usage and token totals
LLM pre-review for sandbox-escalation approvals: an independent-context LLM gate answers sandbox escalation requests before they reach the user, falling back to the user on any failure.