Merge branch 'release/1.11.0' into main

Makefile

@@ -1,5 +1,5 @@
 # =============================================================================
-# Knowledge Genome - Makefile v. 1.10.0
+# Knowledge Genome - Makefile v. 1.11.0
 # Orchestrates the setup and management of the knowledge base.
 # =============================================================================
 

deploy/vm101/n8n-pi-wrap

@@ -18,12 +18,12 @@ case "$cmd" in
     set -a; . "${HOME}/.config/knowledge-genome.env" 2>/dev/null || true; set +a
     exec "${HOME}/.pi/agent/skills/ingest/scripts/orphan-wiki.sh" "$genome"
     ;;
- "pi run")
+  "pi run")
    logger -t n8n-pi-wrap "ok: pi run (prompt via stdin)"
    prompt=$(cat)
    exec /usr/local/bin/pi --no-tools --mode json -p "$prompt" </dev/null
    ;;
- "pi ingest "*)
+  "pi ingest "*)
    # Strict positional parse: EXACTLY `pi ingest <genome> <raw_path>` (two tokens).
    rest="${cmd#pi ingest }"
    genome="${rest%% *}"
@@ -79,7 +79,45 @@ case "$cmd" in
    # MECHANICAL step: validate manifest -> index/log/scoped-lint/commit/PR -> 1 JSON line
    exec "${HOME}/.pi/agent/skills/ingest/scripts/run-ingest.sh" "${genome}"
    ;;
- "pi changed-raw "*)
+  "pi ingest-rework "*)
+    # args: <genome> <raw_path> <feedback_base64>  (3 token).
+    # Feedback in base64 nell'argv: il nodo SSH di n8n non passa stdin, e cosi' i metacaratteri
+    # della review (apici, newline, $(...)) sono neutralizzati.
+    args="${cmd#pi ingest-rework }"
+    genome="${args%% *}"; tmp="${args#* }"
+    raw_path="${tmp%% *}"; fb_b64="${tmp#* }"
+    if [ "$genome" = "$args" ] || [ "$raw_path" = "$tmp" ] || [ -z "$fb_b64" ]; then
+      echo '{"status":"error","reason":"usage: pi ingest-rework <genome> <raw_path> <feedback_b64>"}'; exit 1
+    fi
+    case "$genome"   in ""|*[!a-z0-9-]*)     echo '{"status":"error","reason":"invalid genome"}'; exit 1;; esac
+    case "$raw_path" in raw/*) : ;; *)       echo '{"status":"error","reason":"raw_path must be under raw/"}'; exit 1;; esac
+    case "$raw_path" in *..*|*//*)           echo '{"status":"error","reason":"raw_path traversal"}'; exit 1;; esac
+    case "$raw_path" in *[!A-Za-z0-9._/-]*)  echo '{"status":"error","reason":"raw_path illegal chars"}'; exit 1;; esac
+    case "$fb_b64"   in *[!A-Za-z0-9+/=]*)   echo '{"status":"error","reason":"feedback not base64"}'; exit 1;; esac
+
+    logger -t n8n-pi-wrap "ok: pi ingest-rework ${genome} ${raw_path}"
+    feedback="$(printf '%s' "$fb_b64" | base64 -d 2>/dev/null || true)"
+
+    # lock per-genoma: serializza con gli ingest normali
+    exec 9>"/run/lock/kg-ingest-${genome}.lock" 2>/dev/null || exec 9>"/tmp/kg-ingest-${genome}.lock"
+    if ! flock -n 9; then
+      echo '{"status":"busy","reason":"another ingest is running for this genome","genome":"'"$genome"'"}'; exit 0
+    fi
+    set -a; . "${HOME}/.config/knowledge-genome.env"; set +a
+    cd "${GENOMES_ROOT}/${genome}" || { echo '{"status":"error","reason":"unknown genome"}'; exit 1; }
+    [ -f "$raw_path" ] || { echo '{"status":"error","reason":"raw file not found"}'; exit 1; }
+    : "${KG_LIB_DIR:=${HOME}/knowledge-genome-orchestrator/lib}"
+    source "${KG_LIB_DIR}/clean-start.sh" 2>/dev/null \
+      || { echo '{"status":"error","reason":"clean-start.sh not found"}'; exit 1; }
+    clean_start || { echo '{"status":"error","reason":"clean-start failed"}'; exit 1; }
+    log="$(mktemp -t pi-rework.XXXXXX.log)"
+    INGEST_FEEDBACK="$feedback" \
+      "${HOME}/.pi/agent/skills/ingest/scripts/ingest-semantic.py" "${genome}" "${raw_path}" \
+        >"$log" 2>&1 \
+      || { echo "{\"status\":\"error\",\"stage\":\"semantic\",\"reason\":\"rework failed\",\"log\":\"${log}\"}"; exit 1; }
+    exec "${HOME}/.pi/agent/skills/ingest/scripts/run-ingest.sh" "${genome}"
+    ;;
+  "pi changed-raw "*)
    # List raw/ files changed between two commits, one per line (the webhook payload
    # does NOT include file lists, so vm101's checkout computes the diff itself).
    rest="${cmd#pi changed-raw }"
@@ -119,11 +157,11 @@ case "$cmd" in
        --arg g "$genome" '{status:"ok", genome:$g, count:length, files:.}'
    fi
    ;;
- "ollama list")
+  "ollama list")
    logger -t n8n-pi-wrap "ok: ollama list"
    exec /usr/local/bin/ollama list
    ;;
- "ollama ps")
+  "ollama ps")
    logger -t n8n-pi-wrap "ok: ollama ps"
    exec /usr/local/bin/ollama ps
    ;;

skills/ingest/scripts/ingest-semantic.py

@@ -32,6 +32,7 @@ TIMEOUT    = int(os.environ.get("INGEST_TIMEOUT", "600"))
 # Unset = omit the flag entirely (correct for plain instruct models such as qwen2.5).
 THINK      = os.environ.get("INGEST_THINK")
 TODAY      = datetime.date.today().isoformat()
+FEEDBACK   = os.environ.get("INGEST_FEEDBACK", "").strip()
 
 
 def die(stage, reason):
@@ -217,13 +218,19 @@ def call_model(max_retries=2, base_delay=2.0):
     existing_conc = ", ".join(sorted(existing_concepts)) or "(none yet)"
     prompt = SYSTEM_PROMPT.format(existing_entities=existing_ents, existing_concepts=existing_conc)
 
+    user_content = (
+        ("REVISION REQUESTED BY THE MAINTAINER (address this explicitly):\n"
+        + FEEDBACK + "\n\n") if FEEDBACK else ""
+    ) + (
+        "Source path: " + raw_rel + "\n\n--- SOURCE START ---\n"
+        + source_text + "\n--- SOURCE END ---\n\nReturn the JSON now."
+    )
+
     payload = {
         "model": MODEL,
         "messages": [
             {"role": "system", "content": prompt},
-            {"role": "user", "content":
+            {"role": "user", "content": user_content },
-                "Source path: " + raw_rel + "\n\n--- SOURCE START ---\n"
-                + source_text + "\n--- SOURCE END ---\n\nReturn the JSON now."},
         ],
         "format": SCHEMA,
         "stream": False,

skills/ingest/scripts/run-ingest.sh

@@ -131,6 +131,7 @@ body="$(mktemp)"
 trap 'rm -f "$body"' EXIT  # auto-clean on any exit (success, fail(), or crash)
 
 {
+  echo "<!-- kg:raw=${raw_source} -->" # marker for the rejection loop (invisible in the render)
   echo "## Summary"
   echo "$pr_summary"
   echo ""

templates/pr-description.md

@@ -2,6 +2,19 @@
 
 <!-- One sentence: goal of this session and source processed. -->
 
+<!--
+REVIEW GUIDELINES (write the guideline as the FIRST word of your review):
+  REWORK:  <what to fix>     -> same branch, guided retry
+  RESTART: <why restart>    -> close PR, start over from scratch
+  SPLIT:   <how to split>       -> close PR, reopen as separate branches
+  REJECT:  <why not>           -> close PR, no retry
+  MERGE                          -> approve and merge
+Rules: one concern per directive; be specific to lines/pages; name the principle
+that was violated; describe the DESIRED STATE; avoid saying “do better.”
+-->
+
+Translated with DeepL.com (free version)
+
 ## Pages Created
 
 | Path              | Type                              | Maturity |