AI Knowledge Feed

In a local deployment of Ollama + AnythingLLM, Ollama provides model inference capability, and AnythingLLM handles RAG process orchestration.

nomic-embed-text is an example of an embedding model that can be pulled via Ollama.

Localized RAG deployment (documents do not leave the local environment, models run locally) can keep privacy-sensitive internal materials from having to be uploaded to third parties.

Setting max token in AnythingLLM means setting the number of context windows; the larger the setting, the more accurate the results, but the longer the parsing time.

The knowledge boundary of a general-purpose large model is determined by its training data, so it cannot answer questions related to non-public data such as internal enterprise documents, product prices, and customer information.

In his verification, Mike first asked "What are Gupao's employee reward and punishment policies", and the general-purpose model indicated there was no related content; after uploading the internal policy documents to the AnythingLLM workspace, the system automatically completed text splitting and vectorization and stored them in LanceDB, and when the question was asked again, the model retrieved the knowledge base and answered by citing document clauses.

Large models may need quantization (such as 4-bit quantization) to fit into consumer-grade graphics cards.

Locally deployed open-source models completely bypass the API billing system and consume only local hardware resources.

Online large-model service providers typically authenticate and meter calls via API keys, and bill by the total number of input and output tokens.

Typical application scenarios for Ollama include local model experimentation, private development, and teaching and research.

Exoproc uses a thread hijacking approach: it suspends the target thread, moves the RIP to a sleep stub, and executes code in the existing context.

The 14-byte far relay gate corresponds to the form FF 25 00000000 plus an 8-byte absolute address, and can jump anywhere within the entire 64-bit address space.

Inline hook is an instrumentation approach that overwrites several bytes at the entry of the target function to redirect execution to a custom handler function, and x64 functions need to reserve at least 5 bytes for a JMP rel32 instruction.

Exoproc can handle 5-byte JMP trampolines as well as 14-byte far relay gates.

Exoproc is a stdio-based open-source MCP server that lets Claude Code connect to active Windows x64 processes.