Initial commit: runpod-session.sh with README and CLAUDE.md

Bash script to manage RunPod Ollama pod lifecycle for opencode:
spin up / resume pod, wait for Ollama, patch opencode.jsonc baseURL,
warm up models into VRAM. Includes per-GPU confirmation prompt and
automatic fallback on SUPPLY_CONSTRAINT errors.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

3 files changed, 813 insertions, 0 deletions

diff --git a/CLAUDE.md b/CLAUDE.md
new file mode 100644
index 0000000..4b1b242
--- /dev/null
+++ b/CLAUDE.md
@@ -0,0 +1,51 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## What this is
+
+Single bash script (`runpod-session.sh`) that manages RunPod GPU pod lifecycle for running Ollama models, then patches `~/.config/opencode/opencode.jsonc` so opencode points at the live pod.
+
+## Running / testing
+
+```bash
+# Check syntax
+bash -n runpod-session.sh
+
+# Lint
+shellcheck runpod-session.sh
+
+# Run (requires RUNPOD_API_KEY in ~/.config/runpod-session/config)
+./runpod-session.sh --status
+./runpod-session.sh --model qwen3-coder:latest
+./runpod-session.sh --stop
+./runpod-session.sh --all-models
+./runpod-session.sh --new --gpu-type "RTX PRO 6000" --max-price 1.50
+```
+
+## Architecture
+
+All logic is in `main()` which runs these steps in order:
+
+1. **Existing pod check** — queries RunPod GraphQL API, matches pod by name containing "ollama"
+2. **Pod creation** — if none found: picks cheapest secure-cloud GPU under `MAX_PRICE_PER_HR`, creates pod with `ollama/ollama:latest` image, network volume mounted at `/workspace`
+3. **Wait for Ollama** — polls `https://<pod-id>-11434.proxy.runpod.net/api/tags` until `.models` appears
+4. **Patch opencode.json** — updates `baseURL`, active `model`, and merges `WARMUP_MODELS` into provider models block using jq `*` merge (existing config wins on conflicts)
+5. **Warmup** — POST to `/api/generate` with a dummy prompt to load model into VRAM at `WARMUP_NUM_CTX` context length
+6. **Save state** — writes `~/.config/runpod-session/state.json` with pod_id, url, model, timestamp
+
+## Key design constraints
+
+- All RunPod calls go through `gql()` — single curl wrapper that exits on API errors
+- Pod is identified by name matching `test("ollama"; "i")` — not by ID — so the name `ollama-session` set at creation must not change
+- `patch_opencode_config()` writes a `.bak` before touching opencode.jsonc; jq `*` merge means existing per-model settings survive
+- `OLLAMA_MODELS_PATH` env var on the pod is not set by the script — must be set in config if models live outside default location on the network volume
+- GPU selection only queries `secureCloud == true` pods; community cloud is excluded
+
+## Config and state files
+
+| Path | Purpose |
+|------|---------|
+| `~/.config/runpod-session/config` | Sourced as bash; holds API key, defaults |
+| `~/.config/runpod-session/state.json` | Last session record (pod_id, url, model, timestamp) |
+| `~/.config/opencode/opencode.jsonc` | Patched in-place; `.bak` written before changes |
diff --git a/README.md b/README.md
new file mode 100644
index 0000000..d12d1ec
--- /dev/null
+++ b/README.md
@@ -0,0 +1,154 @@
+# runpod-session.sh
+
+A single bash script that spins up (or resumes) a RunPod GPU pod running Ollama, waits for it to be reachable, warms up your models into VRAM, and patches your `opencode.jsonc` to point at the live pod — all in one command.
+
+## Requirements
+
+- `curl` and `jq`
+- A [RunPod](https://runpod.io) account with:
+  - An API key
+  - A network volume (for persistent model storage)
+- [opencode](https://opencode.ai) installed and configured with a `runpod` provider block
+
+## Installation
+
+```bash
+chmod +x runpod-session.sh
+# optionally symlink to somewhere on your PATH:
+ln -s "$(pwd)/runpod-session.sh" ~/.local/bin/runpod-session
+```
+
+On first run the script creates `~/.config/runpod-session/config` with defaults and exits — edit it, then re-run.
+
+## Configuration
+
+Config file: `~/.config/runpod-session/config`
+
+```bash
+# runpod-session configuration
+
+RUNPOD_API_KEY="rpa_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
+
+# Network volume name exactly as shown in RunPod dashboard
+NETWORK_VOLUME_NAME="my-storage"
+
+# Path where Ollama looks for models on the network volume
+OLLAMA_MODELS_PATH="/workspace/models"
+
+# Must match the key in your opencode.jsonc "provider" block
+OPENCODE_PROVIDER="runpod"
+
+# Model to activate by default (used when --model is not passed)
+DEFAULT_MODEL="qwen3-coder:latest"
+
+# All models stored on this pod — space-separated Ollama tags.
+# These get registered in opencode.jsonc and are warmed up with --all-models.
+WARMUP_MODELS="qwen3-coder:latest translategemma:27b"
+
+# Preferred GPU display name (partial match, case-insensitive). Empty = show all within budget.
+DEFAULT_GPU_TYPE="RTX PRO 6000"
+
+# Hard $/hr ceiling — only GPUs at or below this price are shown
+MAX_PRICE_PER_HR=2.50
+
+# Pod configuration
+CONTAINER_DISK_GB=15
+GPU_COUNT=1
+STARTUP_TIMEOUT=240     # seconds to wait for Ollama to become reachable
+WARMUP_NUM_CTX=32768    # context size used when pre-loading models into VRAM
+```
+
+## Usage
+
+```
+runpod-session.sh [OPTIONS]
+
+Options:
+  --model MODEL        Ollama model tag to warm up (default: DEFAULT_MODEL in config)
+  --all-models         Warm up ALL models listed in WARMUP_MODELS
+  --gpu-type 'NAME'    Preferred GPU display name (partial match, case-insensitive)
+  --max-price PRICE    Max $/hr ceiling (overrides config)
+  --new                Force creation of a new pod (skip restart logic)
+  --stop               Stop the current running pod
+  --status             Show current session state and reachability
+  --help               Show this help
+```
+
+### Typical workflow
+
+```bash
+# Start a session (warm up default model)
+runpod-session.sh
+
+# Start with a specific model
+runpod-session.sh --model qwen3-coder:latest
+
+# Warm up all configured models
+runpod-session.sh --all-models
+
+# Check what's running and how much it's costing
+runpod-session.sh --status
+
+# Stop the pod when done (pod is stopped, not terminated — network volume persists)
+runpod-session.sh --stop
+
+# Force a fresh pod (terminates any existing stopped pod, creates new)
+runpod-session.sh --new --gpu-type "RTX 4090" --max-price 1.80
+```
+
+## How it works
+
+1. **Existing pod check** — queries RunPod for any pod with "ollama" in its name.
+   - If running and reachable: skips straight to warmup.
+   - If stopped/exited: prompts to restart, delete, or abort.
+   - `--new` skips this check entirely.
+
+2. **GPU selection** — queries the RunPod GPU catalog for secure-cloud instances within your price ceiling, sorted by price. If `DEFAULT_GPU_TYPE` is set, matching GPUs are shown first. Each candidate is shown with its VRAM and price; you confirm, skip (`n`), or abort (`a`) for each one. If a GPU you confirm turns out to have no available machines (`SUPPLY_CONSTRAINT`), the script moves to the next candidate automatically.
+
+3. **Pod creation** — deploys `ollama/ollama:latest` on the chosen GPU with your network volume mounted at `/workspace`. Ollama is configured to listen on `0.0.0.0:11434` and keep models in VRAM for 1 hour after last use.
+
+4. **Startup wait** — polls `https://<pod-id>-11434.proxy.runpod.net/api/tags` every 5 seconds until Ollama responds (up to `STARTUP_TIMEOUT` seconds).
+
+5. **opencode.jsonc patch** — updates three fields in your opencode config:
+   - `provider.runpod.options.baseURL` → the live pod URL
+   - `model` → `runpod/<DEFAULT_MODEL>`
+   - `provider.runpod.models` → merges all `WARMUP_MODELS` in (existing per-model settings are preserved via jq recursive merge)
+
+   A `.bak` copy is written before any changes.
+
+6. **Model warmup** — sends a short generation request to load the model into VRAM at `WARMUP_NUM_CTX` context length, so the first real request isn't slow.
+
+7. **State file** — saves pod ID, URL, model, and timestamp to `~/.config/runpod-session/state.json` for use by `--status` and `--stop`.
+
+## opencode provider setup
+
+Your `~/.config/opencode/opencode.jsonc` needs a `runpod` provider block before running the script. The script will fill in the `baseURL` on each session start:
+
+```jsonc
+{
+  "provider": {
+    "runpod": {
+      "options": {
+        "baseURL": ""  // filled in automatically by runpod-session.sh
+      },
+      "models": {}
+    }
+  },
+  "model": "runpod/qwen3-coder:latest"
+}
+```
+
+## Files
+
+| Path | Purpose |
+|------|---------|
+| `~/.config/runpod-session/config` | Main config (sourced as bash) |
+| `~/.config/runpod-session/state.json` | Last session record |
+| `~/.config/opencode/opencode.jsonc` | Patched on each session start |
+| `~/.config/opencode/opencode.jsonc.bak` | Backup written before each patch |
+
+## Cost notes
+
+- **Running pod**: billed at the GPU hourly rate shown during selection.
+- **Stopped pod**: network volume storage continues at ~$0.002/hr — terminate the pod if you no longer need it.
+- `--stop` stops (does not terminate) the pod so it can be quickly restarted without losing the volume.
diff --git a/runpod-session.sh b/runpod-session.sh
new file mode 100755
index 0000000..c81e8dd
--- /dev/null
+++ b/runpod-session.sh
@@ -0,0 +1,608 @@
+#!/usr/bin/env bash
+# runpod-session.sh — Manage RunPod Ollama sessions for opencode
+#
+# Usage:
+#   runpod-session.sh [OPTIONS]
+#
+# Options:
+#   --model MODEL        Ollama model tag to warm up (e.g. qwen3-coder:latest)
+#                        Defaults to DEFAULT_MODEL in config
+#   --all-models         Warm up ALL models listed in WARMUP_MODELS config
+#   --gpu-type 'NAME'    Preferred GPU display name (partial match, case-insensitive)
+#   --max-price PRICE    Max $/hr ceiling (default: MAX_PRICE_PER_HR in config)
+#   --new                Force creation of a new pod (skip restart logic)
+#   --stop               Stop the current running pod
+#   --status             Show current session state and reachability
+#   --help               Show this help
+#
+# Requires: curl, jq
+# Config:   ~/.config/runpod-session/config  (auto-created on first run)
+
+set -euo pipefail
+
+# ─── Paths ────────────────────────────────────────────────────────────────────
+OPENCODE_CONFIG="$HOME/.config/opencode/opencode.jsonc"
+SESSION_CONFIG_DIR="$HOME/.config/runpod-session"
+SESSION_CONFIG="$SESSION_CONFIG_DIR/config"
+SESSION_STATE="$SESSION_CONFIG_DIR/state.json"
+RUNPOD_API="https://api.runpod.io/graphql"
+
+# ─── Defaults (overridden by config file) ─────────────────────────────────────
+OLLAMA_IMAGE="ollama/ollama:latest"
+NETWORK_VOLUME_NAME="danixland-storage"
+OPENCODE_PROVIDER="runpod"
+DEFAULT_MODEL="qwen3-coder:latest"
+WARMUP_MODELS="qwen3-coder:latest translategemma:27b"
+MAX_PRICE_PER_HR=2.50
+CONTAINER_DISK_GB=15
+DEFAULT_GPU_TYPE=""
+GPU_COUNT=1
+POLL_INTERVAL=5
+STARTUP_TIMEOUT=240
+WARMUP_NUM_CTX=32768
+
+# ─── Colors ───────────────────────────────────────────────────────────────────
+RED='\033[0;31m'; YELLOW='\033[1;33m'; GREEN='\033[0;32m'
+CYAN='\033[0;36m'; BOLD='\033[1m'; RESET='\033[0m'
+
+log()  { echo -e "${CYAN}[runpod]${RESET} $*" >&2; }
+ok()   { echo -e "${GREEN}[ok]${RESET} $*" >&2; }
+warn() { echo -e "${YELLOW}[warn]${RESET} $*" >&2; }
+die()  { echo -e "${RED}[error]${RESET} $*" >&2; exit 1; }
+
+# ─── Dependency check ─────────────────────────────────────────────────────────
+for cmd in curl jq; do
+    command -v "$cmd" &>/dev/null || die "Required command not found: $cmd"
+done
+
+# ─── Config bootstrap ─────────────────────────────────────────────────────────
+mkdir -p "$SESSION_CONFIG_DIR"
+
+if [[ ! -f "$SESSION_CONFIG" ]]; then
+    warn "No config found — creating $SESSION_CONFIG"
+    cat > "$SESSION_CONFIG" <<'CONF'
+# runpod-session configuration — edit then re-run.
+
+RUNPOD_API_KEY=""
+
+# Network volume name as shown in RunPod dashboard
+NETWORK_VOLUME_NAME="danixland-storage"
+
+# Must match the key in your opencode.json "provider" block
+OPENCODE_PROVIDER="runpod"
+
+# Model to activate by default (used when --model is not passed)
+DEFAULT_MODEL="qwen3-coder:latest"
+
+# All models that live on this pod — space-separated Ollama tags.
+# These get registered in opencode.json and are warmed up with --all-models.
+WARMUP_MODELS="qwen3-coder:latest translategemma:27b"
+
+# GPU selection
+DEFAULT_GPU_TYPE=""        # e.g. "RTX PRO 6000" — empty = cheapest available
+MAX_PRICE_PER_HR=2.50      # hard $/hr ceiling
+
+# Pod configuration
+CONTAINER_DISK_GB=15
+GPU_COUNT=1
+STARTUP_TIMEOUT=240        # seconds before giving up waiting for Ollama
+WARMUP_NUM_CTX=32768       # num_ctx used when warming up models into VRAM
+CONF
+    echo ""
+    echo -e "${YELLOW}Edit ${BOLD}$SESSION_CONFIG${RESET}${YELLOW}, set RUNPOD_API_KEY, then re-run.${RESET}"
+    exit 0
+fi
+
+# shellcheck source=/dev/null
+source "$SESSION_CONFIG"
+[[ -z "${RUNPOD_API_KEY:-}" ]] && die "RUNPOD_API_KEY not set in $SESSION_CONFIG"
+
+# ─── Argument parsing ─────────────────────────────────────────────────────────
+OPT_MODEL="${DEFAULT_MODEL}"
+OPT_ALL_MODELS=0
+OPT_GPU_TYPE="${DEFAULT_GPU_TYPE:-}"
+OPT_MAX_PRICE="${MAX_PRICE_PER_HR}"
+OPT_FORCE_NEW=0
+OPT_STOP=0
+OPT_STATUS=0
+
+while [[ $# -gt 0 ]]; do
+    case "$1" in
+        --model)       OPT_MODEL="$2";      shift 2 ;;
+        --all-models)  OPT_ALL_MODELS=1;    shift   ;;
+        --gpu-type)    OPT_GPU_TYPE="$2";   shift 2 ;;
+        --max-price)   OPT_MAX_PRICE="$2";  shift 2 ;;
+        --new)         OPT_FORCE_NEW=1;     shift   ;;
+        --stop)        OPT_STOP=1;          shift   ;;
+        --status)      OPT_STATUS=1;        shift   ;;
+        --help|-h)     sed -n '2,17p' "$0"; exit 0  ;;
+        *) die "Unknown option: $1  (use --help)" ;;
+    esac
+done
+
+# ─── RunPod GraphQL helper ────────────────────────────────────────────────────
+gql() {
+    local err
+    local response
+    response=$(curl --ipv4 -s \
+        -H "Content-Type: application/json" \
+        -H "Authorization: Bearer $RUNPOD_API_KEY" \
+        -d "$1" \
+        "$RUNPOD_API")
+    if [[ $? -ne 0 || -z "$response" ]]; then
+        die "RunPod API request failed — check API key and connectivity"
+    fi
+    err=$(echo "$response" | jq -r '.errors[0].message // empty' 2>/dev/null || true)
+    [[ -n "$err" ]] && die "RunPod API error: $err"
+    echo "$response"
+}
+
+# ─── Pod queries ──────────────────────────────────────────────────────────────
+get_pods() {
+    gql '{"query":"{ myself { pods { id name desiredStatus costPerHr runtime { uptimeInSeconds } machine { gpuDisplayName } } } }"}'
+}
+
+find_ollama_pod() {
+    # Returns compact JSON of the first pod matching "ollama" in its name, or empty string
+    local result
+    result=$(echo "$1" | jq -c \
+        '.data.myself.pods[] | select(.name | test("ollama"; "i"))' 2>/dev/null | head -1)
+    echo "$result"
+}
+
+# ─── --status subcommand ──────────────────────────────────────────────────────
+cmd_status() {
+    echo ""
+    echo -e "${BOLD}runpod-session status${RESET}"
+    echo ""
+
+    # ── Live pod data from API ────────────────────────────────────────────────
+    local pods_json pod_json
+    pods_json=$(get_pods)
+    pod_json=$(find_ollama_pod "$pods_json")
+
+    if [[ -n "$pod_json" ]]; then
+        local pod_id status gpu cost_hr uptime_sec cost_so_far
+        pod_id=$(echo "$pod_json"     | jq -r '.id')
+        status=$(echo "$pod_json"     | jq -r '.desiredStatus')
+        gpu=$(echo "$pod_json"        | jq -r '.machine.gpuDisplayName // "?"')
+        cost_hr=$(echo "$pod_json"    | jq -r '.costPerHr // 0')
+        uptime_sec=$(echo "$pod_json" | jq -r '.runtime.uptimeInSeconds // 0')
+
+        # Calculate cost: (uptime_seconds / 3600) * cost_per_hr
+        cost_so_far=$(echo "$uptime_sec $cost_hr" | awk '{printf "%.4f", ($1/3600)*$2}')
+        uptime_human=$(echo "$uptime_sec" | awk '{
+            h=int($1/3600); m=int(($1%3600)/60); s=$1%60
+            if (h>0) printf "%dh %dm %ds", h, m, s
+            else if (m>0) printf "%dm %ds", m, s
+            else printf "%ds", s
+        }')
+
+        echo -e "  Pod:      ${BOLD}${pod_id}${RESET}"
+        echo -e "  GPU:      ${gpu}"
+        echo -e "  Status:   ${status}"
+        echo -e "  Rate:     \$${cost_hr}/hr"
+        if [[ "$status" == "RUNNING" ]]; then
+            echo -e "  Uptime:   ${uptime_human}"
+            echo -e "  Cost:     \$${cost_so_far} this session"
+        else
+            echo -e "  Uptime:   —"
+            echo -e "  Cost:     —"
+        fi
+        echo ""
+
+        # ── Ollama reachability ───────────────────────────────────────────────
+        local base="https://${pod_id}-11434.proxy.runpod.net"
+        local tags_response
+        tags_response=$(curl -s --ipv4 --max-time 5 "${base}/api/tags" 2>/dev/null || true)
+        if echo "$tags_response" | jq -e '.models' > /dev/null 2>&1; then
+            echo -e "  Ollama:   ${GREEN}reachable${RESET}"
+            local models
+            models=$(echo "$tags_response" | jq -r '.models[].name' 2>/dev/null || true)
+            if [[ -n "$models" ]]; then
+                echo -e "  In VRAM:  $(echo "$models" | tr '\n' ' ')"
+            else
+                echo -e "  In VRAM:  none (model will load on first request)"
+            fi
+        else
+            echo -e "  Ollama:   ${YELLOW}not reachable${RESET}"
+        fi
+    else
+        echo -e "  ${YELLOW}No Ollama pod found in your account.${RESET}"
+    fi
+
+    # ── Saved session state ───────────────────────────────────────────────────
+    if [[ -f "$SESSION_STATE" ]]; then
+        echo ""
+        echo -e "  ${BOLD}Last session record:${RESET}"
+        jq -r '"  Started:  \(.started_at)\n  Model:    \(.model)\n  URL:      \(.ollama_url)"' \
+            "$SESSION_STATE"
+    fi
+
+    echo ""
+}
+
+# ─── Pod lifecycle ────────────────────────────────────────────────────────────
+restart_pod() {
+    log "Restarting pod $1 ..."
+    gql "{\"query\":\"mutation { podResume(input: { podId: \\\"$1\\\", gpuCount: $GPU_COUNT }) { id desiredStatus } }\"}" > /dev/null
+}
+
+stop_pod() {
+    log "Stopping pod $1 ..."
+    gql "{\"query\":\"mutation { podStop(input: { podId: \\\"$1\\\" }) { id desiredStatus } }\"}" > /dev/null
+    ok "Pod stopped. Storage costs continue while stopped (~\$0.002/hr)."
+}
+
+terminate_pod() {
+    log "Terminating pod $1 ..."
+    gql "{\"query\":\"mutation { podTerminate(input: { podId: \\\"$1\\\" }) }\"}" > /dev/null
+    ok "Pod terminated."
+}
+
+# ─── --stop subcommand ────────────────────────────────────────────────────────
+cmd_stop() {
+    local pod_id=""
+    if [[ -f "$SESSION_STATE" ]]; then
+        pod_id=$(jq -r '.pod_id' "$SESSION_STATE")
+    else
+        local pods_json pod_json
+        pods_json=$(get_pods)
+        pod_json=$(find_ollama_pod "$pods_json")
+        [[ -z "$pod_json" ]] && die "No Ollama pod found in your account."
+        pod_id=$(echo "$pod_json" | jq -r '.id')
+    fi
+    echo -n "  Stop pod $pod_id? [y/N] "
+    read -r confirm
+    [[ "${confirm,,}" != "y" ]] && { log "Aborted."; exit 0; }
+    stop_pod "$pod_id"
+    rm -f "$SESSION_STATE"
+}
+
+# ─── GPU selection ────────────────────────────────────────────────────────────
+# Returns a JSON array of candidates sorted by price (preferred first if set).
+# Caller iterates and tries each until one succeeds.
+get_gpu_candidates() {
+    local preferred="$1" max_price="$2"
+    log "Querying available GPUs (secure cloud, max \$$max_price/hr) ..." >&2
+
+    local result candidates
+    result=$(curl -s --ipv4 \
+        -H "Content-Type: application/json" \
+        -H "Authorization: Bearer $RUNPOD_API_KEY" \
+        -d '{"query":"{ gpuTypes { id displayName memoryInGb secureCloud lowestPrice(input: { gpuCount: 1 }) { uninterruptablePrice } } }"}' \
+        "$RUNPOD_API")
+
+    candidates=$(echo "$result" | jq -c \
+        --arg max "$max_price" \
+        '[.data.gpuTypes[]
+          | select(.secureCloud == true)
+          | select((.lowestPrice.uninterruptablePrice // 0) > 0)
+          | select(.lowestPrice.uninterruptablePrice <= ($max | tonumber))
+          | { id: .id, name: .displayName, vram: .memoryInGb,
+              price: .lowestPrice.uninterruptablePrice }
+         ] | sort_by(.price)')
+
+    [[ -z "$candidates" || "$candidates" == "[]" ]] && \
+        die "No GPUs on secure cloud within \$$max_price/hr. Try --max-price."
+
+    # Bubble preferred GPU to front so it's tried first
+    if [[ -n "$preferred" ]]; then
+        local reordered
+        reordered=$(echo "$candidates" | jq -c \
+            --arg p "$preferred" \
+            '( [.[] | select(.name | test($p; "i"))] ) +
+             ( [.[] | select(.name | test($p; "i") | not)] )')
+        local pcount
+        pcount=$(echo "$reordered" | jq 'map(select(.name | test($p; "i"))) | length' --arg p "$preferred" 2>/dev/null || echo 0)
+        if [[ "$pcount" -eq 0 ]]; then
+            warn "Preferred GPU '$preferred' not in catalog within price limit. Will try all." >&2
+        fi
+        candidates="$reordered"
+    fi
+
+    echo "$candidates"
+}
+
+# ─── Network volume ───────────────────────────────────────────────────────────
+get_network_volume_id() {
+    local result vol_id
+    result=$(curl -s --ipv4 \
+        -H "Content-Type: application/json" \
+        -H "Authorization: Bearer $RUNPOD_API_KEY" \
+        -d '{"query":"{ myself { networkVolumes { id name } } }"}' \
+        "$RUNPOD_API")
+    vol_id=$(echo "$result" | jq -r \
+        --arg n "$NETWORK_VOLUME_NAME" \
+        '.data.myself.networkVolumes[] | select(.name == $n) | .id')
+    [[ -z "$vol_id" ]] && die "Network volume '$NETWORK_VOLUME_NAME' not found."
+    echo "$vol_id"
+}
+
+# ─── Create pod ───────────────────────────────────────────────────────────────
+create_pod() {
+    local gpu_id="$1" vol_id="$2"
+    log "Creating pod ..."
+
+    # Build env array — omit OLLAMA_MODELS entry if path is unset
+    local env_json
+    env_json='[{"key":"OLLAMA_HOST","value":"0.0.0.0"},{"key":"OLLAMA_LOAD_TIMEOUT","value":"10m"},{"key":"OLLAMA_KEEP_ALIVE","value":"1h"}]'
+    if [[ -n "${OLLAMA_MODELS_PATH:-}" ]]; then
+        env_json=$(echo "$env_json" | jq \
+            --arg v "$OLLAMA_MODELS_PATH" \
+            '. + [{"key":"OLLAMA_MODELS","value":$v}]')
+    fi
+
+    local payload
+    payload=$(jq -n \
+        --arg gpu_id    "$gpu_id" \
+        --arg vol_id    "$vol_id" \
+        --arg image     "$OLLAMA_IMAGE" \
+        --argjson gpu_count   "$GPU_COUNT" \
+        --argjson disk        "$CONTAINER_DISK_GB" \
+        --argjson env         "$env_json" \
+        '{query: "mutation($input: PodFindAndDeployOnDemandInput!) { podFindAndDeployOnDemand(input: $input) { id } }",
+          variables: { input: {
+            cloudType: "SECURE",
+            gpuCount: $gpu_count,
+            volumeInGb: 0,
+            containerDiskInGb: $disk,
+            minVcpuCount: 4,
+            minMemoryInGb: 15,
+            gpuTypeId: $gpu_id,
+            name: "ollama-session",
+            imageName: $image,
+            ports: "11434/http",
+            volumeMountPath: "/workspace",
+            networkVolumeId: $vol_id,
+            env: $env
+          }}}')
+
+    local result pod_id err_code
+    result=$(curl -s --ipv4 \
+        -H "Content-Type: application/json" \
+        -H "Authorization: Bearer $RUNPOD_API_KEY" \
+        -d "$payload" \
+        "$RUNPOD_API")
+
+    pod_id=$(echo "$result" | jq -r '.data.podFindAndDeployOnDemand.id // empty')
+    if [[ -n "$pod_id" ]]; then
+        echo "$pod_id"
+        return 0
+    fi
+
+    err_code=$(echo "$result" | jq -r '.errors[0].extensions.code // empty')
+    if [[ "$err_code" == "SUPPLY_CONSTRAINT" ]]; then
+        warn "No supply for GPU $gpu_id — trying next candidate ..." >&2
+        return 1
+    fi
+    die "Pod creation failed. Response: $result"
+}
+
+# ─── Wait for Ollama ──────────────────────────────────────────────────────────
+wait_for_pod() {
+    local pod_id="$1"
+    local url="https://${pod_id}-11434.proxy.runpod.net"
+    local elapsed=0
+    log "Polling $url/api/tags (timeout: ${STARTUP_TIMEOUT}s) ..."
+    while (( elapsed < STARTUP_TIMEOUT )); do
+        local response
+        response=$(curl -s --ipv4 --max-time 5 "${url}/api/tags" 2>/dev/null || true)
+        if echo "$response" | jq -e '.models' > /dev/null 2>&1; then
+            echo ""; ok "Ollama is up."; return 0
+        fi
+        printf "  [%3ds] waiting...\r" "$elapsed"
+        sleep "$POLL_INTERVAL"
+        (( elapsed += POLL_INTERVAL ))
+    done
+    echo ""
+    die "Timed out after ${STARTUP_TIMEOUT}s. Check the RunPod dashboard."
+}
+
+# ─── Patch opencode.json ──────────────────────────────────────────────────────
+#
+# What this changes in your opencode.json:
+#   .provider.runpod.options.baseURL     → https://<pod-id>-11434.proxy.runpod.net/v1
+#   .model                               → runpod/qwen3-coder:latest  (DEFAULT_MODEL)
+#   .provider.runpod.models              → merges all WARMUP_MODELS in (preserving
+#                                          any existing per-model config you have)
+#
+# Uses jq's `*` (recursive merge) so your existing model overrides are never clobbered.
+# A .bak backup is written before any changes.
+#
+patch_opencode_config() {
+    local new_url="$1"   # full URL including /v1
+
+    [[ ! -f "$OPENCODE_CONFIG" ]] && die "opencode config not found at $OPENCODE_CONFIG"
+    cp "$OPENCODE_CONFIG" "${OPENCODE_CONFIG}.bak"
+
+    # Build a jq object for all warmup models: { "model:tag": {"tools":true}, ... }
+    # tools:true is a safe default — it won't override existing per-model settings
+    # because we merge with * where existing config wins on conflicts.
+    local models_patch="{}"
+    for m in $WARMUP_MODELS; do
+        models_patch=$(printf '%s' "$models_patch" \
+            | jq --arg m "$m" '. + {($m): {"tools": true}}')
+    done
+
+    local tmp
+    tmp=$(mktemp)
+    jq \
+        --arg provider  "$OPENCODE_PROVIDER" \
+        --arg url       "$new_url" \
+        --arg model     "${OPENCODE_PROVIDER}/${DEFAULT_MODEL}" \
+        --argjson patch "$models_patch" \
+        '
+        .provider[$provider].options.baseURL = $url
+        | .model = $model
+        | .provider[$provider].models = (
+            $patch * (.provider[$provider].models // {})
+          )
+        ' "$OPENCODE_CONFIG" > "$tmp" && mv "$tmp" "$OPENCODE_CONFIG"
+
+    ok "opencode.json patched:"
+    log "  provider.${OPENCODE_PROVIDER}.options.baseURL = $new_url"
+    log "  model = ${OPENCODE_PROVIDER}/${DEFAULT_MODEL}"
+}
+
+# ─── Warm up one model ────────────────────────────────────────────────────────
+warmup_model() {
+    local pod_id="$1" model="$2"
+    local base="https://${pod_id}-11434.proxy.runpod.net"
+
+    log "Warming up '$model' into VRAM ..."
+
+    if curl -s --ipv4 --max-time 300 -X POST "${base}/api/generate" \
+        -H "Content-Type: application/json" \
+        -d "{\"model\": \"$model\", \"prompt\": \"hi\", \"stream\": false, \"options\": {\"num_ctx\": ${WARMUP_NUM_CTX:-32768}}}" \
+        > /dev/null 2>&1; then
+        ok "  '$model' is loaded."
+    else
+        warn "  Warmup for '$model' failed — model will load on first use."
+    fi
+}
+
+# ─── State persistence ────────────────────────────────────────────────────────
+save_state() {
+    jq -n \
+        --arg pod_id "$1" \
+        --arg url    "$2" \
+        --arg model  "$3" \
+        --arg ts     "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
+        '{ pod_id: $pod_id, ollama_url: $url, model: $model, started_at: $ts }' \
+        > "$SESSION_STATE"
+}
+
+# ─── Main ─────────────────────────────────────────────────────────────────────
+main() {
+    [[ $OPT_STATUS -eq 1 ]] && { cmd_status; exit 0; }
+    [[ $OPT_STOP   -eq 1 ]] && { cmd_stop;   exit 0; }
+
+    echo -e "${BOLD}runpod-session${RESET} — Ollama on RunPod → opencode"
+    echo ""
+
+    local pod_id=""
+    local _skip_wait=0
+
+    # ── 1. Existing pod check ─────────────────────────────────────────────────
+    if [[ $OPT_FORCE_NEW -eq 0 ]]; then
+        log "Checking for existing Ollama pods ..."
+        local pods_json pod_json
+        pods_json=$(get_pods)
+        pod_json=$(find_ollama_pod "$pods_json")
+
+        if [[ -n "$pod_json" ]]; then
+            pod_id=$(echo "$pod_json" | jq -r '.id')
+            local status gpu cost
+            status=$(echo "$pod_json" | jq -r '.desiredStatus')
+            gpu=$(echo "$pod_json"    | jq -r '.machine.gpuDisplayName // "?"')
+            cost=$(echo "$pod_json"   | jq -r '.costPerHr // "?"')
+
+            echo -e "  Found: ${BOLD}${pod_id}${RESET}  GPU: ${gpu}  \$${cost}/hr  Status: ${status}"
+            echo ""
+
+            case "$status" in
+                RUNNING)
+                    local _check
+                    _check=$(curl -s --ipv4 --max-time 5                         "https://${pod_id}-11434.proxy.runpod.net/api/tags" 2>/dev/null || true)
+                    if echo "$_check" | jq -e '.models' > /dev/null 2>&1; then
+                        ok "Already running and reachable — skipping startup sequence."
+                        _skip_wait=1
+                    else
+                        log "Pod is running but Ollama not yet reachable — waiting ..."
+                    fi
+                    ;;
+                EXITED|STOPPED)
+                    echo -n "  [R]estart  [D]elete and create new  [A]bort  [R/d/a]: "
+                    read -r choice
+                    case "${choice,,}" in
+                        d) terminate_pod "$pod_id"; pod_id="" ;;
+                        a) log "Aborted."; exit 0 ;;
+                        *) restart_pod "$pod_id" ;;
+                    esac
+                    ;;
+                *)
+                    warn "Unexpected pod state '$status' — ignoring this pod."
+                    pod_id=""
+                    ;;
+            esac
+        else
+            log "No existing Ollama pod found."
+        fi
+    fi
+
+    # ── 2. Create new pod if needed ───────────────────────────────────────────
+    if [[ -z "$pod_id" ]]; then
+        local vol_id candidates gpu_json gpu_id gpu_name gpu_vram gpu_price
+        vol_id=$(get_network_volume_id)
+        ok "Volume: ${NETWORK_VOLUME_NAME} ($vol_id)"
+
+        candidates=$(get_gpu_candidates "$OPT_GPU_TYPE" "$OPT_MAX_PRICE")
+        local count
+        count=$(echo "$candidates" | jq 'length')
+        [[ "$count" -eq 0 ]] && die "No GPU candidates found."
+
+        log "${count} GPU options within budget. Will prompt for each."
+
+        local i=0
+        while [[ $i -lt $count ]]; do
+            gpu_json=$(echo "$candidates" | jq -c --argjson i "$i" '.[$i]')
+            gpu_id=$(echo "$gpu_json"    | jq -r '.id')
+            gpu_name=$(echo "$gpu_json"  | jq -r '.name')
+            gpu_vram=$(echo "$gpu_json"  | jq -r '.vram')
+            gpu_price=$(echo "$gpu_json" | jq -r '.price')
+
+            echo ""
+            echo -e "  ${BOLD}${gpu_name}${RESET}  ${gpu_vram}GB VRAM  \$${gpu_price}/hr"
+            echo -n "  Create pod with this GPU? [Y/n/a(bort)] "
+            read -r choice
+            case "${choice,,}" in
+                a) log "Aborted."; exit 0 ;;
+                n) (( i++ )) || true; continue ;;
+            esac
+
+            if pod_id=$(create_pod "$gpu_id" "$vol_id"); then
+                ok "Pod created: $pod_id  GPU: ${gpu_name}"
+                break
+            fi
+            (( i++ )) || true
+        done
+
+        [[ -z "$pod_id" ]] && die "All ${count} GPU candidates exhausted. Try --max-price or later."
+    fi
+
+    # ── 3. Wait for Ollama if needed ─────────────────────────────────────────
+    if [[ "${_skip_wait:-0}" != "1" ]]; then
+        wait_for_pod "$pod_id"
+    fi
+
+    # ── 4. Final URL ──────────────────────────────────────────────────────────
+    local ollama_url="https://${pod_id}-11434.proxy.runpod.net/v1"
+
+    # ── 5. Patch opencode.json ────────────────────────────────────────────────
+    patch_opencode_config "$ollama_url"
+
+    # ── 6. Warmup ─────────────────────────────────────────────────────────────
+    if [[ $OPT_ALL_MODELS -eq 1 ]]; then
+        for m in $WARMUP_MODELS; do
+            warmup_model "$pod_id" "$m"
+        done
+    elif [[ -n "$OPT_MODEL" ]]; then
+        warmup_model "$pod_id" "$OPT_MODEL"
+    else
+        warn "No warmup requested. Use --model MODEL or --all-models."
+    fi
+
+    # ── 7. Save state ─────────────────────────────────────────────────────────
+    save_state "$pod_id" "$ollama_url" "$OPT_MODEL"
+
+    # ── 8. Done ───────────────────────────────────────────────────────────────
+    echo ""
+    echo -e "${BOLD}${GREEN}Ready.${RESET}"
+    printf "  Pod:   %s\n" "$pod_id"
+    printf "  URL:   %s\n" "$ollama_url"
+    printf "  Run:   opencode\n"
+    echo ""
+}
+
+main "$@"