Skip to content

Solution for long term memory for agent coding CLIs and to facilitate handoff between different agent vendors

Rust

docs: state the 8350-token consolidation floor in the default config and error

AkitaOnRailsaa299d573ff1
diff --git a/crates/ai-memory-cli/src/config.rs b/crates/ai-memory-cli/src/config.rs
index 2f0be448..76d10d43 100644
--- a/crates/ai-memory-cli/src/config.rs
+++ b/crates/ai-memory-cli/src/config.rs
@@ -2099,7 +2099,8 @@ impl Config {
if effective_input_chars < required_input_chars {
anyhow::bail!(
"consolidation.max_input_tokens must leave at least {required_input_chars} \
- characters for the prompt (got {}, which is {effective_input_chars} at \
+ characters for the prompt ({min_input_tokens} tokens at the default \
+ input_token_safety_margin) (got {}, which is {effective_input_chars} at \
input_token_safety_margin = {safety_margin}); below that the system \
prompt and page conventions leave no room for observations. Raise \
max_input_tokens, or raise input_token_safety_margin towards 1.0",
diff --git a/crates/ai-memory-cli/templates/config.default.toml b/crates/ai-memory-cli/templates/config.default.toml
index bb1ce097..a68b0063 100644
--- a/crates/ai-memory-cli/templates/config.default.toml
+++ b/crates/ai-memory-cli/templates/config.default.toml
@@ -176,7 +176,9 @@ per_user = false
# LLM consolidation prompt limits. These are approximate because tokenizers
# differ by provider and model. Their sum must fit the model's context window;
# leave additional headroom for tokenizer variance. Small local models must
-# lower both values together. Startup minimums are 6500 input and 1000 output.
+# lower both values together. Startup minimums are 8350 input (at the default
+# input_token_safety_margin of 0.8; a lower margin needs proportionally more)
+# and 1000 output.
# Consolidation turns a finished session's raw observations into wiki
# pages. Off (zero-LLM) you still get session pages + search; with an
# LLM you get distilled concept/gotcha/rule pages. Reach for these knobs
diff --git a/crates/ai-memory-store/src/writer.rs b/crates/ai-memory-store/src/writer.rs
index 1df105d8..3d9710ba 100644
--- a/crates/ai-memory-store/src/writer.rs
+++ b/crates/ai-memory-store/src/writer.rs
@@ -1673,9 +1673,10 @@ impl WriterHandle {
rx.await.map_err(|_| StoreError::WriterClosed)?
}
- /// Persist one open-session checkpoint job (PreCompact / PostCompaction)
- /// for the current observation generation. Duplicate generations are
- /// idempotent. An already-ended session is refused.
+ /// Persist one checkpoint job (PreCompact / PostCompaction) for the
+ /// current observation generation. Duplicate generations are idempotent.
+ /// The session may already have ended: a resumed session keeps
+ /// `ended_at` while it captures again.
pub async fn enqueue_session_checkpoint(
&self,
workspace_id: WorkspaceId,
@@ -1693,7 +1694,8 @@ impl WriterHandle {
rx.await.map_err(|_| StoreError::WriterClosed)?
}
- /// Atomically claim the oldest due SessionEnd consolidation job.
+ /// Atomically claim the oldest due session consolidation job (SessionEnd
+ /// or checkpoint).
pub async fn claim_session_consolidation(
&self,
now: i64,