ComfyUI-Session-Memory

🧠 ComfyUI-Session-Memory

Structured workflow memory and human-in-the-loop creative feedback for ComfyUI.

ComfyUI-Session-Memory gives your generation workflows persistent memory across runs. It stores both the raw prompts and structured summaries, injects summaries into your LLM system prompt for context-aware generation, and provides an interactive in-node feedback panel where you can rate and annotate every prompt — all without leaving ComfyUI.

---

✨ Features

• Session Memory Reader — Loads the accumulated session history and outputs the summaries for injection into any LLM system prompt.
• Session Memory Writer — Receives both the raw prompt and a structured summary, storing them as paired entries in the session's persistent JSON file.
• Session Feedback Editor — Interactive in-node UI to rate (⭐ 1–5) and annotate each prompt. Displays the raw prompt text (not the summary) for accurate creative review. Supports visual thumbnail previews of generated images.
• Feedback injection — Rated prompts are automatically converted into REPLICATE / AVOID directives injected into the next run's system prompt, closing the creative feedback loop.
• Thumbnail system — Accepts a full image batch, downscales to ≤1 MP, and saves them organized as sessions/{id}/run_{N}/prompt_{P}/img_{K}.jpg. Supports multi-pass batches (batch_count > 1).

---

🔄 Architecture

🧠 Reader ──session_memory──► LLM System Prompt ──► Sampler
    ↑                                                   │
    │                                              💾 Writer
    │                                                   │
    │                                          ⭐ Feedback Editor
    │                                                   │
    └────────────── sessions/{session_id}.json ◄────────┘

The Reader emits the accumulated summaries before the current run begins. The Writer stores both the prompt and summary after the run completes. The Feedback Editor displays the prompt text for review — it never shows the summary.

---

📦 Nodes

🧠 Session Memory Reader

Loads and outputs the full accumulated session log.

| Input | Type | Description | |---|---|---| | session_id | STRING | Session identifier (must match Writer/Editor) | | reset_session | BOOLEAN | Clears all memory and thumbnails for this session | | max_runs | INT | Sliding window: keep only the last N runs (0 = unlimited) |

| Output | Type | Description | |---|---|---| | session_memory | STRING | Full log for injection into system prompt | | session_feedback | STRING | REPLICATE/AVOID directives from rated runs | | run_count | INT | Number of runs recorded in this session |

---

💾 Session Memory Writer

Stores both the raw prompt and a structured summary as a paired entry.

| Input | Type | Description | |---|---|---| | session_id | STRING | Session identifier | | prompt | STRING | The raw prompt sent to the model. Displayed in the Feedback Editor | | summary | STRING | Structured summary from a Gemini/LLM summarizer. Returned by the Reader | | max_runs | INT | Sliding window: keep only the last N runs (0 = unlimited) |

| Output | Type | Description | |---|---|---| | run_summary | STRING | Pass-through of the summary | | session_id | STRING | Pass-through for chaining to Feedback Editor |

---

⭐ Session Feedback Editor

Interactive in-node panel for post-run creative review.

| Input | Type | Description | |---|---|---| | session_id | STRING | Session identifier | | mode | ENUM | non_blocking (annotate later) or blocking (pause until saved) | | num_prompts | INT | Number of prompts in the run | | batch_count | INT | Sampling passes per prompt (ComfyUI's batch count) | | run_summary | STRING | Optional — connect from Writer to enforce execution order | | thumbnail_images | IMAGE | Optional — full image batch for thumbnail generation |

UI features (inside the node):

• Run navigator (← Run 3/5 →) to browse any past run
• Per-prompt star rating (1–5 ⭐)
• Per-prompt text notes
• Thumbnail grid — side-by-side images for each batch pass
• Save Feedback button — persists ratings/notes to the session JSON

---

📁 File Structure

ComfyUI-Session-Memory/
├── session_memory.py       # All node logic + API routes
├── __init__.py             # Node registration
├── web/
│   └── session_feedback_editor.js  # Frontend extension (LiteGraph)
└── sessions/               # Auto-created at runtime (gitignored)
    ├── my_project.json     # Session data: entries + feedback
    └── my_project/
        └── run_5/
            ├── prompt_1/
            │   ├── img_0.jpg
            │   └── img_1.jpg   ← second batch pass
            └── prompt_2/
                ├── img_0.jpg
                └── img_1.jpg

---

🚀 Installation

Option A — Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/Latentnaut/ComfyUI-Session-Memory

Restart ComfyUI.

Option B — ComfyUI Manager: Search for Session Memory in the ComfyUI Manager and install.

---

💡 Recommended Workflow

1. Reader → inject session_memory + session_feedback into your LLM system prompt.
2. LLM generates structured prompts aware of past runs.
3. Sampler generates images (connect the batch output to Feedback Editor thumbnail_images).
4. A Gemini/LLM summarizer creates a run_summary → send to Writer.
5. Feedback Editor receives the run_summary (for ordering) and the image batch.
6. Rate and annotate prompts inside the node. Click Save Feedback.
7. On the next Run, the Reader injects the feedback as REPLICATE/AVOID directives automatically.

---

⚙️ Thumbnail Details

• Images are downscaled to a maximum of 1 megapixel (LANCZOS) and saved as JPEG (quality 85).
• Supports batch_count > 1: ComfyUI re-executes the graph once per batch pass, and thumbnails are appended (not overwritten) on each pass.
• Distribution: image i in the batch → prompt_{i % num_prompts}, pass index auto-incremented.

---

📋 Session JSON Format

{
  "run_count": 5,
  "entries": [
    {
      "prompt": "--- RUN 1 ---\nPROMPT 1: concept: urban decay...\n",
      "summary": "--- RUN 1 ---\nSummary of urban decay concept...\n"
    }
  ],
  "feedback": {
    "3": {
      "prompts": {
        "1": { "rating": 5, "notes": "Perfect composition" },
        "2": { "rating": 1, "notes": "Too repetitive" }
      }
    }
  }
}

Backwards compatibility: Legacy sessions with plain-string entries are automatically migrated on load. Each old string becomes both the prompt and summary.

---

📄 License

MIT