LLM Wiki
An LLM wiki is a shared markdown knowledge base a model reads from and edits across sessions.
What is an LLM wiki?
An LLM wiki is a set of markdown files a model reads at the start of a session, edits as it works, and leaves the next session to continue. It works like a personal wiki: one file per topic, links between related files, written by the model instead of a person.
How is an LLM wiki different from a context repository?
An LLM wiki is written by the session that last edited it, without checks on duplicates or contradictions, and no control over file read access: two sessions can each add a note about the same customer under a different name, and nothing merges them. A context repository resolves that: it merges the same entity found in two places into one record, applies permissions at the point of retrieval, and keeps a version history, so many agents across many sessions add to one governed body of knowledge instead of one file changing unpredictably from session to session.
How does an LLM wiki work?
Each session reads the files left by the last one and continues from there, trusting whatever a previous session wrote since nothing marks a note as reviewed or current. Qontext runs a version of this at company scale, with the merging, permissions, and version history from the previous section applying across every agent and every session.