Cereja Flamejante
A knowledge management and agent collaboration system built on the LLM Wiki pattern. It comprises three core components: Núcleo, Editorial Engine and Agentic Factory. It is used to organize knowledge, prepare content for publication and maintain agent roles.
Timeline (last 90 days)
The Cereja Flamejante project implemented the LLM Wiki approach, and its architecture includes components such as Núcleo (recording knowledge, evidence, interpretations, and arguments), Editorial Engine (preparing content for publication), and Agentic Factory (maintaining agent roles and shared controls).
Cereja's ingest command computes the file's SHA-256 hash and incorporates it into the note's filename; if the note already exists, it returns its path and preserves existing annotations.
In Cereja, the current ingest command only creates source-provenance records; document extraction and summarization are not yet implemented.
The working wiki uses the normalized source address (recorded in resource) as the identity identifier; in Cereja, the hash identifies the content and the citation records the source, which allows different byte contents to be detected but does not solve all duplication problems.
In Cereja, the public note option --public-note confirms that the filename and metadata may appear publicly, but does not grant any rights to the document itself; raw files in raw/ are by default not included in Git; before publishing a PDF its license agreement must be checked.
Cereja's notes are created in draft status, recording the title, declared author, raw file path, hash value, URL when available, declared rights status, and ingest date, with unknown fields marked as unknown.
In Cereja, the checks for coverage, rights, and promotion still need adaptation and testing, while also respecting the status definitions already in use by Núcleo.
All Facts (7)
The Cereja Flamejante project implemented the LLM Wiki approach, and its architecture includes components such as Núcleo (recording knowledge, evidence, interpretations, and arguments), Editorial Engine (preparing content for publication), and Agentic Factory (maintaining agent roles and shared controls).
50%UnverifiedCereja's ingest command computes the file's SHA-256 hash and incorporates it into the note's filename; if the note already exists, it returns its path and preserves existing annotations.
50%UnverifiedIn Cereja, the current ingest command only creates source-provenance records; document extraction and summarization are not yet implemented.
50%UnverifiedThe working wiki uses the normalized source address (recorded in resource) as the identity identifier; in Cereja, the hash identifies the content and the citation records the source, which allows different byte contents to be detected but does not solve all duplication problems.
50%UnverifiedIn Cereja, the public note option --public-note confirms that the filename and metadata may appear publicly, but does not grant any rights to the document itself; raw files in raw/ are by default not included in Git; before publishing a PDF its license agreement must be checked.
50%UnverifiedCereja's notes are created in draft status, recording the title, declared author, raw file path, hash value, URL when available, declared rights status, and ingest date, with unknown fields marked as unknown.
50%UnverifiedIn Cereja, the checks for coverage, rights, and promotion still need adaptation and testing, while also respecting the status definitions already in use by Núcleo.
50%