Deterministic Reasoning Engine (DRE) for dsh: dre__* tools for knowledge verification (three-stage discrimination), deterministic cognition loops, constraint solving, mental models and synapse memory - strengthens information certainty.
Install
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:ListenJ/axiom-dre-dsh
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
Axiom Deterministic Reasoning Engine (DRE) as a DeepSeek Harness (dsh) plugin.
A self-contained plugin: it bundles the DRE engine and its MCP backend, and exposes the engine's tools to dsh under the dre__ prefix — knowledge verification (three-stage discrimination), deterministic cognition loops, constraint solving, mental models and synapse memory.
- License: MIT
- Hot-pluggable:
dsh plugin add/rm - Runtime: requires Bun
Install
No extra setup — the plugin carries its own backend:
dsh plugin --profile web add github:ListenJ/axiom-dre-dsh
Restart dsh (dsh web). The tool list then includes the dre__* tools and the dre_plugin_status diagnostic tool.
Uninstall
dsh plugin --profile web rm axiom-dre-dsh
Architecture
The plugin is self-contained: it bundles the DRE engine and a DRE-only MCP backend (backend/server.js, a single-file Bun build) and spawns it over stdio:
dsh (Node) ── axiom-dre-dsh ──stdio──▶ built-in backend (Bun) ──▶ DRE engine
│ │
filter + register Kernel / Pipeline / …
- The plugin bundles the DRE engine (Kernel, three-stage verification pipeline, cognitive loop, constraint solver, synapse memory) and spawns it with
bun backend/server.js --stdio; a writabledata/directory is created automatically.
dre_plugin_status (always available)
A diagnostic tool reporting MCP bridge state (connected, tool count, server name), DRE engine status, and the effective config summary.
Configuration
Overridden via cordis.patch.yml under line id dre (overriding the whole section replaces all keys).
| Key | Default | Description |
|---|---|---|
dataDir |
<plugin>/data |
Backend data directory (SQLite / memory). Created automatically. |
mcpEnabled |
true |
Launch and bridge the backend. |
mcpCommand / mcpArgs |
bun / <plugin>/backend/server.js --stdio |
Backend launch command. |
mcpServerName |
dre |
Public tool prefix (<serverName>__<tool>). |
mcpToolCallTimeoutMs |
60000 |
Per-tool call timeout (ms). |
mcpFailOnStartupError |
false |
false = tolerate startup failure (warn only); true = fail the fiber on initial connect error. |
toolFilter |
[] |
Empty = built-in DRE allow-list; explicit array fully replaces it (prefix ending _ or exact name). |
synapseEnabled |
true |
When false, drops mind_synapse_* and mind_suggest. |
Tools (prefix dre__)
- Knowledge verification:
dre_write_knowledgedre_search_knowledgedre_read_knowledgedre_subgraphdre_statusdre_constraint_inject - Deterministic cognition:
cognitive_loopcognitive_loop_fullcognitive_pipeline_runcognitive_pipeline_run_fullcognitive_statetask_graph_execute - Reasoning graph:
reasoning_buildreasoning_detect_gapsreasoning_fill_gapreasoning_result - Constraint solving:
constraint_checkconstraint_listconstraint_select_bestconstraint_stats - Actor / mental model:
actor_listactor_sendmental_model_listmental_model_matchmental_model_predict - Synapse memory:
mind_suggestmind_synapse_activatemind_synapse_createmind_synapse_spreadmind_synapse_suggestmind_synapse_tracemind_synapse_verify - Diagnostic:
dre_plugin_status
License
MIT
Links
More in this category
Tencent/WeKnora#dsh-weknora★ 34028
Four read-only tools over a WeKnora knowledge base: list knowledge bases, hybrid passage search, reassemble one document's chunks in order, and WeKnora's own cited RAG or ReAct-agent answer with a resumable session id.
superdesigndev/treg★ 5298
Tool catalog for agents: search ~2,600 external endpoints (SEO and SERP, backlinks, social, people and company enrichment, ad libraries, scraping) by the task you want done, read each one's parameters and per-call price, then call it with the credential injected server-side. Ships the skill plus an MCP row that stays disabled until TREG_TOKEN is set.
TencentCloudBase/CloudBase-AI-Toolkit#dsh-plugin★ 1136
Tencent CloudBase backend for DeepSeek Harness — scaffold and deploy full-stack apps from chat, render query results as table cards with paging, sorting and CSV export, preview a deployment on its domain, and call the CloudBase MCP toolset (`mcp__cloudbase__*`) with device-code login.
gitroomhq/postiz-agent#dsh-postiz★ 517
Connects DeepSeek Harness to Postiz over MCP: list connected social media channels, fetch per-platform posting rules, and schedule, draft, or publish posts to X, LinkedIn, Instagram, Facebook, Threads, TikTok, YouTube, Reddit, Bluesky, Mastodon, Discord, Slack, Telegram and more; adds a postiz workflow skill.
EthanYoQ/Invoice-Downloader#dsh-invoice-downloader★ 507
Local IMAP invoice download, OCR, archive, and Excel reimbursement summaries for DeepSeek Harness.
anysearch-team/anysearch-dsh★ 461
AnySearch-powered real-time web and vertical search provider for DeepSeek Harness.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.