Cross-agent doctor and the graph-write path it exposed
By RepoJournal , from @anurag629's public GitHub activity
Backfilled
1 person shipped this
Anurag added a canary test that proved one agent could write data and another could recall it from a shared dataset, but the test immediately caught a critical gap in the write path.
The work landed in anurag629/cortex-bridge as a canary that demonstrates the full premise of cross-agent memory before building on it. The test writes a session QA from one agent and recalls it as a different agent on the same pinned dataset [1]. The first implementation tried to use improve() to bridge raw session QA into the graph, following the same pattern the plugin uses today, but it failed: improve(session_ids) returns in under a second and does not cognify raw session QA into the graph at all [1]. The test showed that cross-agent memory cannot ride on improve alone. In response, Anurag added client.ingestToGraph, a new method that takes node_set and a wait option, and a passing doctor running on the local Ollama server [2]. No plugin regressions resulted from the change [2].
One email a day. Unsubscribe in one click.
A short briefing every day anurag629 ships something — in about 3 minutes.
One email a day. Unsubscribe in one click. Read a past issue →
References
- [1] Add a cross-agent doctor, and the graph-write path it proved we need ↗ anurag629/cortex-bridge
- [2] Add a cross-agent doctor, and the graph-write path it proved we need anurag629/cortex-bridge
