@nbardy/oompa 0.3.1 → 0.4.1

package/README.md

@@ -99,7 +99,7 @@ This spawns:
 
 | Field | Required | Description |
 |-------|----------|-------------|
-| `model` | yes | `harness:model` (e.g. `claude:opus-4.5`, `codex:codex-5.2-mini`) |
+| `model` | yes | `harness:model` or `harness:model:reasoning` (e.g. `codex:o3:low`, `claude:opus-4.6`) |
 | `prompt` | no | String or array of paths — concatenated into one prompt |
 | `iterations` | no | Max iterations per worker (default: 10) |
 | `count` | no | Number of workers with this config (default: 1) |

package/agentnet/src/agentnet/cli.clj

@@ -134,6 +134,43 @@
 
 (declare cmd-swarm)
 
+(defn- probe-model
+  "Send 'say ok' to a model via its harness CLI. Returns true if model responds."
+  [harness model]
+  (try
+    (let [cmd (case harness
+                :claude ["claude" "--model" model "-p" "say ok" "--max-turns" "1"]
+                :codex  ["codex" "exec" "--model" model "--" "say ok"])
+          result (process/sh cmd {:out :string :err :string :timeout 30000})]
+      (zero? (:exit result)))
+    (catch Exception _ false)))
+
+(defn- validate-models!
+  "Probe each unique harness:model pair. Prints results and exits if any fail."
+  [worker-configs review-model]
+  (let [models (cond-> (set (map (fn [wc]
+                                   (parse-model-string (:model wc)))
+                                 worker-configs))
+                 review-model (conj review-model))
+        _ (println "Validating models...")
+        results (pmap (fn [{:keys [harness model]}]
+                        (let [ok (probe-model harness model)]
+                          (println (format "  %s:%s %s"
+                                           (name harness) model
+                                           (if ok "OK" "FAIL")))
+                          {:harness harness :model model :ok ok}))
+                      models)
+        failures (filter (complement :ok) results)]
+    (when (seq failures)
+      (println)
+      (println "ERROR: The following models are not accessible:")
+      (doseq [{:keys [harness model]} failures]
+        (println (format "  %s:%s" (name harness) model)))
+      (println)
+      (println "Fix model names in oompa.json and retry.")
+      (System/exit 1))
+    (println)))
+
 (defn cmd-run
   "Run orchestrator — uses oompa.json if present, otherwise simple mode"
   [opts args]
@@ -284,11 +321,15 @@
                        (if available? "✓ available" "✗ not found"))))))
 
 (defn- parse-model-string
-  "Parse 'harness:model' string into {:harness :model}"
+  "Parse model string into {:harness :model :reasoning}.
+   Formats: 'harness:model', 'harness:model:reasoning', or just 'model'."
   [s]
   (if (and s (str/includes? s ":"))
-    (let [[h m] (str/split s #":" 2)]
-      {:harness (keyword h) :model m})
+    (let [parts (str/split s #":" 3)]
+      (case (count parts)
+        2 {:harness (keyword (first parts)) :model (second parts)}
+        3 {:harness (keyword (first parts)) :model (second parts) :reasoning (nth parts 2)}
+        {:harness :codex :model s}))
     {:harness :codex :model s}))
 
 (defn cmd-swarm
@@ -324,12 +365,13 @@
           ;; Convert to worker format
           workers (map-indexed
                     (fn [idx wc]
-                      (let [{:keys [harness model]} (parse-model-string (:model wc))]
+                      (let [{:keys [harness model reasoning]} (parse-model-string (:model wc))]
                         (worker/create-worker
                           {:id (str "w" idx)
                            :swarm-id swarm-id
                            :harness harness
                            :model model
+                           :reasoning reasoning
                            :iterations (or (:iterations wc) 10)
                            :prompts (:prompt wc)
                            :can-plan (:can_plan wc)
@@ -352,6 +394,9 @@
                            (if (:prompt wc) (str ", " (:prompt wc)) "")))))
       (println)
 
+      ;; Preflight: probe each unique model before launching workers
+      (validate-models! worker-configs review-model)
+
       ;; Run workers using new worker module
       (worker/run-workers! workers))))
 

package/agentnet/src/agentnet/worker.clj

@@ -106,8 +106,9 @@
 (defn create-worker
   "Create a worker config.
    :prompts is a string or vector of strings — paths to prompt files.
-   :can-plan when false, worker waits for tasks before starting (backpressure)."
-  [{:keys [id swarm-id harness model iterations prompts can-plan review-harness review-model]}]
+   :can-plan when false, worker waits for tasks before starting (backpressure).
+   :reasoning reasoning effort level (e.g. \"low\", \"medium\", \"high\") — codex only."
+  [{:keys [id swarm-id harness model iterations prompts can-plan reasoning review-harness review-model]}]
   {:id id
    :swarm-id swarm-id
    :harness (or harness :codex)
@@ -118,6 +119,7 @@
               (string? prompts) [prompts]
               :else [])
    :can-plan (if (some? can-plan) can-plan true)
+   :reasoning reasoning
    :review-harness review-harness
    :review-model review-model
    :completed 0
@@ -144,7 +146,7 @@
 
 (defn- run-agent!
   "Run agent with prompt, return {:output string, :done? bool, :exit int}"
-  [{:keys [id swarm-id harness model prompts]} worktree-path context]
+  [{:keys [id swarm-id harness model prompts reasoning]} worktree-path context]
   (let [;; 1. Task header (always, from package)
         task-header (or (load-prompt "config/prompts/_task_header.md") "")
 
@@ -175,6 +177,7 @@
                               "--skip-git-repo-check"
                               "-C" abs-worktree]
                        model (into ["--model" model])
+                       reasoning (into ["-c" (str "reasoning_effort=\"" reasoning "\"")])
                        true (conj "--" full-prompt))
               :claude (cond-> ["claude" "-p" "--dangerously-skip-permissions"
                                "--session-id" session-id]

package/bin/test-models

@@ -0,0 +1,130 @@
+#!/usr/bin/env bash
+# test-models — end-to-end validation of models in oompa.json
+#
+# Usage: test-models [path/to/oompa.json]
+#
+# For each unique model, launches the agent and asks it to write a result
+# file. Then checks all expected files exist. This validates the full
+# pipeline: harness CLI → model access → code execution → file I/O.
+
+set -euo pipefail
+
+CONFIG="${1:-oompa.json}"
+
+if [ ! -f "$CONFIG" ]; then
+  echo "Config not found: $CONFIG"
+  echo "Usage: test-models [path/to/oompa.json]"
+  exit 1
+fi
+
+# Extract unique model strings from workers[] and review_model
+MODELS=$(python3 -c "
+import json
+with open('$CONFIG') as f:
+    cfg = json.load(f)
+models = set()
+if cfg.get('review_model'):
+    models.add(cfg['review_model'])
+for w in cfg.get('workers', []):
+    if 'model' in w:
+        models.add(w['model'])
+for m in sorted(models):
+    print(m)
+")
+
+if [ -z "$MODELS" ]; then
+  echo "No models found in $CONFIG"
+  exit 1
+fi
+
+# Create results directory
+RUN_ID=$(python3 -c "import uuid; print(str(uuid.uuid4())[:8])")
+RESULTS_DIR="tst_results_${RUN_ID}"
+mkdir -p "$RESULTS_DIR"
+
+MODEL_COUNT=$(echo "$MODELS" | wc -l | tr -d ' ')
+echo "Testing $MODEL_COUNT models from $CONFIG"
+echo "Results dir: $RESULTS_DIR"
+echo ""
+
+# Launch all models in parallel
+PIDS=()
+MODEL_NAMES=()
+
+while IFS= read -r model; do
+  HARNESS="${model%%:*}"
+  # Strip reasoning suffix for the model name passed to CLI
+  REST="${model#*:}"
+  MODEL_NAME="${REST%%:*}"
+  # Safe filename: replace slashes and dots
+  SAFE_NAME=$(echo "$model" | tr '/:.' '_')
+
+  MODEL_NAMES+=("$SAFE_NAME")
+  PROMPT="Write a file called ${RESULTS_DIR}/${SAFE_NAME}_DONE with exactly the text DONE. Nothing else. Just create that one file."
+
+  echo "  launching $model ..."
+
+  case "$HARNESS" in
+    claude)
+      claude --model "$MODEL_NAME" -p "$PROMPT" --dangerously-skip-permissions --max-turns 3 \
+        > "${RESULTS_DIR}/${SAFE_NAME}.log" 2>&1 &
+      PIDS+=($!)
+      ;;
+    codex)
+      codex exec --model "$MODEL_NAME" \
+        --dangerously-bypass-approvals-and-sandbox \
+        --skip-git-repo-check \
+        -- "$PROMPT" \
+        > "${RESULTS_DIR}/${SAFE_NAME}.log" 2>&1 &
+      PIDS+=($!)
+      ;;
+    *)
+      echo "    SKIP (unknown harness: $HARNESS)"
+      # Remove from expected list
+      unset 'MODEL_NAMES[${#MODEL_NAMES[@]}-1]'
+      ;;
+  esac
+done <<< "$MODELS"
+
+# Wait for all
+echo ""
+echo "Waiting for all models to complete..."
+for pid in "${PIDS[@]}"; do
+  wait "$pid" 2>/dev/null || true
+done
+
+# Check results
+echo ""
+echo "Results:"
+echo ""
+
+PASS=0
+FAIL=0
+
+for safe_name in "${MODEL_NAMES[@]}"; do
+  RESULT_FILE="${RESULTS_DIR}/${safe_name}_DONE"
+  printf "  %-40s " "$safe_name"
+
+  if [ -f "$RESULT_FILE" ]; then
+    CONTENT=$(cat "$RESULT_FILE" | tr -d '[:space:]')
+    if [ "$CONTENT" = "DONE" ]; then
+      echo "PASS"
+      PASS=$((PASS + 1))
+    else
+      echo "FAIL (file exists but content: '$(head -1 "$RESULT_FILE")')"
+      FAIL=$((FAIL + 1))
+    fi
+  else
+    echo "FAIL (no result file)"
+    # Show first few lines of log for debugging
+    if [ -f "${RESULTS_DIR}/${safe_name}.log" ]; then
+      head -5 "${RESULTS_DIR}/${safe_name}.log" | sed 's/^/    /'
+    fi
+    FAIL=$((FAIL + 1))
+  fi
+done
+
+echo ""
+echo "$PASS passed, $FAIL failed (results in $RESULTS_DIR/)"
+
+[ "$FAIL" -eq 0 ]

package/package.json

@@ -1,11 +1,12 @@
 {
   "name": "@nbardy/oompa",
-  "version": "0.3.1",
+  "version": "0.4.1",
   "description": "Git-worktree multi-agent swarm orchestrator for Codex and Claude",
   "license": "MIT",
   "type": "commonjs",
   "bin": {
-    "oompa": "bin/oompa.js"
+    "oompa": "bin/oompa.js",
+    "oompa-test-models": "bin/test-models"
   },
   "files": [
     "bin/",