@nbardy/oompa 0.4.0 → 0.4.2

package/README.md

@@ -82,24 +82,22 @@ This repo has a fleshed out version of the idea. The oompa loompas are organized
 **oompa.json** — the only file you need:
 ```json
 {
-  "review_model": "codex:codex-5.2",
   "workers": [
-    {"model": "claude:opus-4.5", "prompt": ["config/prompts/planner.md"], "iterations": 5, "count": 1},
-    {"model": "codex:codex-5.2-mini", "prompt": ["config/prompts/executor.md"], "iterations": 10, "count": 3, "can_plan": false}
+    {"model": "claude:opus", "prompt": ["config/prompts/planner.md"], "iterations": 5, "count": 1},
+    {"model": "codex:gpt-5.3-codex:medium", "prompt": ["config/prompts/executor.md"], "iterations": 10, "count": 3, "can_plan": false}
   ]
 }
 ```
 
 This spawns:
 - **1 planner** (opus) — reads spec, explores codebase, creates/refines tasks
-- **3 executors** (mini) — claims and executes tasks fast
-- **Reviews** done by codex-5.2 before any merge
+- **3 executors** (gpt-5.3-codex, medium reasoning) — claims and executes tasks fast
 
 #### Worker fields
 
 | Field | Required | Description |
 |-------|----------|-------------|
-| `model` | yes | `harness:model` or `harness:model:reasoning` (e.g. `codex:o3:low`, `claude:opus-4.6`) |
+| `model` | yes | `harness:model` or `harness:model:reasoning` (e.g. `codex:gpt-5.3-codex:medium`, `claude:opus`) |
 | `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

@@ -16,6 +16,7 @@
             [agentnet.worker :as worker]
             [agentnet.tasks :as tasks]
             [agentnet.agent :as agent]
+            [babashka.process :as process]
             [clojure.string :as str]
             [clojure.java.io :as io]
             [cheshire.core :as json]))
@@ -132,15 +133,15 @@
 ;; Commands
 ;; =============================================================================
 
-(declare cmd-swarm)
+(declare cmd-swarm parse-model-string)
 
 (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"])
+                :claude ["claude" "--model" model "-p" "say ok"]
+                :codex  ["codex" "exec" "--dangerously-bypass-approvals-and-sandbox" "--skip-git-repo-check" "--model" model "--" "say ok"])
           result (process/sh cmd {:out :string :err :string :timeout 30000})]
       (zero? (:exit result)))
     (catch Exception _ false)))
@@ -343,10 +344,9 @@
       (println)
       (println "Create oompa.json with format:")
       (println "{")
-      (println "  \"review_model\": \"codex:codex-5.2\",")
       (println "  \"workers\": [")
-      (println "    {\"model\": \"codex:codex-5.2-mini\", \"prompt\": \"prompts/executor.md\", \"iterations\": 10, \"count\": 3, \"can_plan\": false},")
-      (println "    {\"model\": \"claude:opus-4.5\", \"prompt\": [\"prompts/base.md\", \"prompts/planner.md\"], \"count\": 1}")
+      (println "    {\"model\": \"codex:gpt-5.3-codex:medium\", \"prompt\": \"prompts/executor.md\", \"iterations\": 10, \"count\": 3, \"can_plan\": false},")
+      (println "    {\"model\": \"claude:opus\", \"prompt\": [\"prompts/base.md\", \"prompts/planner.md\"], \"count\": 1}")
       (println "  ]")
       (println "}")
       (println)
@@ -449,7 +449,7 @@
   (println "  --keep-worktrees         Don't cleanup worktrees after run")
   (println)
   (println "Examples:")
-  (println "  ./swarm.bb loop 10 --harness codex --model codex-5.2-mini --workers 3")
+  (println "  ./swarm.bb loop 10 --harness codex --model gpt-5.3-codex --workers 3")
   (println "  ./swarm.bb loop --workers claude:5 codex:4 --iterations 20")
   (println "  ./swarm.bb swarm oompa.json  # Run multi-model config"))
 

package/agentnet/src/agentnet/worker.clj

@@ -177,7 +177,7 @@
                               "--skip-git-repo-check"
                               "-C" abs-worktree]
                        model (into ["--model" model])
-                       reasoning (into ["-c" (str "reasoning_effort=\"" reasoning "\"")])
+                       reasoning (into ["-c" (str "model_reasoning_effort=\"" reasoning "\"")])
                        true (conj "--" full-prompt))
               :claude (cond-> ["claude" "-p" "--dangerously-skip-permissions"
                                "--session-id" session-id]

package/bin/test-models

@@ -1,10 +1,11 @@
 #!/usr/bin/env bash
-# test-models — probe all models in oompa.json with a hello-world check
+# test-models — end-to-end validation of models in oompa.json
 #
 # Usage: test-models [path/to/oompa.json]
 #
-# Sends "say ok" to each unique model via its harness CLI.
-# Reports pass/fail for each. Exits non-zero if any fail.
+# 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
 
@@ -16,9 +17,9 @@ if [ ! -f "$CONFIG" ]; then
   exit 1
 fi
 
-# Extract unique model strings (harness:model) from workers[] and review_model
+# Extract unique model strings from workers[] and review_model
 MODELS=$(python3 -c "
-import json, sys
+import json
 with open('$CONFIG') as f:
     cfg = json.load(f)
 models = set()
@@ -36,43 +37,94 @@ if [ -z "$MODELS" ]; then
   exit 1
 fi
 
-echo "Probing models from $CONFIG"
+# Create results directory
+RUN_ID=$(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 ""
 
-PASS=0
-FAIL=0
+# Launch all models in parallel
+PIDS=()
+MODEL_NAMES=()
 
 while IFS= read -r model; do
   HARNESS="${model%%:*}"
-  MODEL_NAME="${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."
 
-  printf "  %-30s " "$model"
+  echo "  launching $model ..."
 
   case "$HARNESS" in
     claude)
-      OUTPUT=$(claude --model "$MODEL_NAME" -p "say ok" --max-turns 1 2>&1) && EXIT=$? || EXIT=$?
+      claude --model "$MODEL_NAME" -p "$PROMPT" --dangerously-skip-permissions --max-turns 3 \
+        > "${RESULTS_DIR}/${SAFE_NAME}.log" 2>&1 &
+      PIDS+=($!)
       ;;
     codex)
-      OUTPUT=$(codex exec --model "$MODEL_NAME" -- "say ok" 2>&1) && EXIT=$? || EXIT=$?
+      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)"
-      continue
+      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
 
-  if [ $EXIT -eq 0 ]; then
-    echo "OK"
-    PASS=$((PASS + 1))
+# 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"
-    # Print first line of error for context
-    echo "$OUTPUT" | head -3 | sed 's/^/    /'
+    echo "FAIL (no result file)"
+    # Show first few lines of log for debugging
+    if [ -f "${RESULTS_DIR}/${safe_name}.log" ]; then
+      head -5 "${RESULTS_DIR}/${safe_name}.log" | sed 's/^/    /'
+    fi
     FAIL=$((FAIL + 1))
   fi
-done <<< "$MODELS"
+done
 
 echo ""
-echo "$PASS passed, $FAIL failed"
+echo "$PASS passed, $FAIL failed (results in $RESULTS_DIR/)"
 
 [ "$FAIL" -eq 0 ]

package/oompa.example.json

@@ -1,15 +1,13 @@
 {
-  "review_model": "codex:codex-5.2",
-
   "workers": [
     {
-      "model": "claude:opus-4.5",
+      "model": "claude:opus",
       "prompt": ["config/prompts/planner.md"],
       "iterations": 5,
       "count": 1
     },
     {
-      "model": "codex:codex-5.2-mini",
+      "model": "codex:gpt-5.3-codex:medium",
       "prompt": ["config/prompts/executor.md"],
       "iterations": 10,
       "count": 3,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nbardy/oompa",
-  "version": "0.4.0",
+  "version": "0.4.2",
   "description": "Git-worktree multi-agent swarm orchestrator for Codex and Claude",
   "license": "MIT",
   "type": "commonjs",