dsh-infinite-context
Multi-tier memory management, dynamic context compaction, and mid-thinking guards for DeepSeek Harness.
Install
$ npm pack && dsh plugin --profile <name> add ./dsh-infinite-context-0.1.0.tgzPlugin Overview & Capabilities
AI-assisted organization based on the public repository snapshot. The content must be grounded in source evidence and does not replace compatibility or security verification.
Key Capabilities
- Progressive token-pressure compression with model-aware dynamic thresholds
- Mid-thinking guard that interrupts LLM stream on real window exhaustion to compact and retry
- Three-tier memory pyramid (short, mid, long) persisted with SQLite
- Semantic RAG retrieval with active context surface deduplication
- Tool-result archiving and instant distillation with recoverable content references
- Reversible compaction preserving folded message ranges
- Comprehensive suite of 11 inspection and maintenance tools (status, probe, search, expand, etc.)
Useful For
- Preventing context exhaustion during long-running agentic tasks and deep-thinking models
- Running agent sessions on local small models with limited context windows
- Maintaining cross-session persistent structured memory and project knowledge
- Auditing and inspecting model context usage, compaction savings, and cache hit rates
Who It Fits
- DeepSeek Harness developers working on complex multi-turn workflows
- Users running local LLMs via Ollama, llama-server, or LM Studio seeking optimal context throughput
- Engineers requiring persistent agent memory and recoverable archived tool outputs
Documented Limitations
- Model probing is limited to supported server interfaces such as Ollama, llama-server, and OpenAI-compatible endpoints
- Uses lightweight feature-hash embeddings by default unless optional HuggingFace Transformers dependencies are configured
- Folded message history and tool archives are subject to SQLite retention day and entry limits
DSH Compatibility
Version-specific runtime evidence collected by DSH Plugin. A missing result means we have not tested that combination yet.
Security Signals
Objective signals discovered from package metadata and source inspection. These are not a guarantee that a plugin is safe.
package.json declares 0 runtime, 3 development, 10 peer, and 0 optional dependencies.
package.json declares DSH bundle metadata.
GitHub reports the repository license as MIT.
A root package.json was captured and can be inspected by the registry.
package.json declares a prepare lifecycle script that may run during relevant package installation workflows.
Public GitHub source metadata is available for this registry snapshot.
Source & Registry Notes
Public provenance, Registry classification, and the latest source check for this entry, kept separate from runtime verification.
- Source repository
- chocobo77/dsh-infinite-context
- Registry source
- GitHub · dsh-plugin topic
- Registry classification
- Plugin Bundle
- Source checked
- 1988fae · 2026-10-17
This project is independently indexed from public source information. DSH Plugin is not affiliated with DeepSeek or the plugin author. Always check the author repository before installation.
Repository Activity
- GitHub stars
- 3GitHub stars
- Forks
- 0
- Open issues
- 0
- Last commit
- 2026-10-03
- Last release
- No release detected
Related DSH Plugins
Ranked by overlapping capabilities, use cases, plugin type, categories, and DSH profile.
Knowledge graph memory and context engine for DeepSeek Harness with cross-session recall.
View plugin →Lingshu long-term auditable memory system and cognitive graph for DeepSeek Harness.
View plugin →Composable, view-based layered memory bundle for DeepSeek Harness with pluggable providers and strategies.
View plugin →LLM-supervised persistent memory and graph-based recall integration bundle for DeepSeek Harness.
View plugin →