add call_simple for non-agent LLM calls

audit, digest, and compare now go through the API backend via
call_simple(), which logs to llm-logs/{caller}/.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

poc-memory/src/agents/audit.rs

@@ -3,7 +3,7 @@
 // Each batch of links gets reviewed by Sonnet, which returns per-link actions:
 // KEEP, DELETE, RETARGET, WEAKEN, STRENGTHEN. Batches run in parallel via rayon.
 
-use super::llm::call_sonnet;
+use super::llm;
 use crate::store::{self, Store, new_relation};
 
 use std::collections::HashSet;
@@ -211,7 +211,7 @@ pub fn link_audit(store: &mut Store, apply: bool) -> Result<AuditStats, String>
     // Run batches in parallel via rayon
     let batch_results: Vec<_> = batch_data.par_iter()
         .map(|(batch_idx, batch_infos, prompt)| {
-            let response = call_sonnet("audit", prompt);
+            let response = llm::call_simple("audit", prompt);
             let completed = done.fetch_add(1, Ordering::Relaxed) + 1;
             eprint!("\r  Batches: {}/{} done", completed, total_batches);
             (*batch_idx, batch_infos, response)

poc-memory/src/agents/digest.rs

@@ -5,7 +5,7 @@
 // summarize weeklies. All three share the same generate/auto-detect
 // pipeline, parameterized by DigestLevel.
 
-use super::llm::{call_sonnet, semantic_keys};
+use super::llm;
 use crate::store::{self, Store, new_relation};
 use crate::neuro;
 
@@ -211,7 +211,7 @@ fn generate_digest(
     }
     println!("  {} inputs", inputs.len());
 
-    let keys = semantic_keys(store);
+    let keys = llm::semantic_keys(store);
     let keys_text = keys.iter()
         .map(|k| format!("  - {}", k))
         .collect::<Vec<_>>()
@@ -244,7 +244,7 @@ fn generate_digest(
     println!("  Prompt: {} chars (~{} tokens)", prompt.len(), prompt.len() / 4);
 
     println!("  Calling Sonnet...");
-    let digest = call_sonnet("digest", &prompt)?;
+    let digest = llm::call_simple("digest", &prompt)?;
 
     let key = digest_node_key(level.name, label);
     store.upsert_provenance(&key, &digest, "digest:write")?;

poc-memory/src/agents/llm.rs

@@ -183,9 +183,26 @@ pub(crate) fn call_haiku(agent: &str, prompt: &str) -> Result<String, String> {
     call_model(agent, "haiku", prompt)
 }
 
-/// Call a model using an agent definition's model and tool configuration.
+/// Simple LLM call for non-agent uses (audit, digest, compare).
-/// Uses the direct API backend when api_base_url is configured,
+/// Logs to llm-logs/{caller}/ file.
-/// otherwise falls back to claude CLI subprocess.
+pub(crate) fn call_simple(caller: &str, prompt: &str) -> Result<String, String> {
+    let log_dir = crate::store::memory_dir().join("llm-logs").join(caller);
+    fs::create_dir_all(&log_dir).ok();
+    let log_path = log_dir.join(format!("{}.txt", crate::store::compact_timestamp()));
+
+    use std::io::Write;
+    let log = move |msg: &str| {
+        if let Ok(mut f) = fs::OpenOptions::new()
+            .create(true).append(true).open(&log_path)
+        {
+            let _ = writeln!(f, "{}", msg);
+        }
+    };
+
+    super::api::call_api_with_tools_sync(caller, prompt, &log)
+}
+
+/// Call a model using an agent definition's configuration.
 pub(crate) fn call_for_def(
     def: &super::defs::AgentDef,
     prompt: &str,

poc-memory/src/cli/agent.rs

@@ -384,11 +384,8 @@ fn llm_compare(
 ) -> Result<std::cmp::Ordering, String> {
     let prompt = build_compare_prompt(a, b);
 
-    let response = if model == "haiku" {
+    let _ = model; // model selection handled by API backend config
-        llm::call_haiku("compare", &prompt)?
+    let response = llm::call_simple("compare", &prompt)?;
-    } else {
-        llm::call_sonnet("compare", &prompt)?
-    };
     let response = response.trim().to_uppercase();
 
     if response.contains("BETTER: B") {