# Newcclm complete description (for AI assistant)

This document is for AI assistants to read and is used to help users configure and troubleshoot Newcclm.
More detailed than /llms.txt, including complete parameters, directly executable examples, and error tables.

---

## 1. Service Overview

Newcclm is an AI assistant service that provides two ways of use:

1. **Command line client** (recommended for ordinary users)
Run `npx newcclm` and log in with your account. No API Key is required, and no configuration is required.

2. **Access your own tools according to the OpenAI compatible protocol**
If you need a key starting with `sk-`, apply to the service provider. Available for Cursor, WorkBuddy,
DeepSeek Harness and other tools that support custom interfaces.

---

## 2. Access parameters (used in method 2)

| Item | Value |
| --- | --- |
| Provider type | Custom Provider / OpenAI Compatible |
| Base URL | `https://sccclm.com/en/v1` |
| API Key | `sk-...` (apply to the service provider) |
| Model | `newcclm-fast` or `newcclm-coding` |
| Protocol | OpenAI compatible (`/v1/chat/completions`) |
| Authentication header | `Authorization: Bearer <API_KEY>` |
| Streaming | Support, pass `"stream": true` |
| Context | 1M token |

# # # Gear Instructions

| Model name | Scenario-oriented | Underlying model |
| --- | --- | --- |
| `newcclm-fast` | Fast file: reply to emails, modify copy, organize information, and ask daily questions | GLM-5.3-Flash |
| `newcclm-coding` | Enhanced files: complex analysis, long documents, code writing | Qwen3.8-Flash |

The backup line is DeepSeek V4-Flash, which automatically takes over only when the main line fails (disaster recovery, non-degradation).
The three underlying models are all 1M contexts.

# # # Public description of the underlying model

This service exposes the underlying model and does not perform hidden replacements:

- **GLM-5.3-Flash** - daily main work, corresponding to `newcclm` / `newcclm-fast`
- **Qwen3.8-Flash** - encoding and long context, corresponding to `newcclm-coding`
- **DeepSeek V4-Flash** —— Backup line, takes over when the main line fails

**Users must also abide by the terms of use of each of the above models**, including but not limited to: not used to generate illegal content,
The model itself may not be resold in its capacity. This is a requirement of the upstream clause for end users (flow-down) and is not exempted by transit.

### Important restrictions

- **Only model names in the table above can be used. ** Fill in `gpt-4o`, `claude-3-5-sonnet`, `deepseek-chat`
will return an error that the model does not exist.
- **Base URL do not overlay `/v1`**, nor `/chat/completions` (unless the client
Complete endpoints are explicitly required).
- **Do not select the Anthropic native protocol**, the connection will fail.

---

## 3. Directly executable verification request

```bash
curl https://sccclm.com/en/v1/chat/completions \
  -H "Authorization: Bearer $NEWCCLM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "newcclm-fast",
"messages": [{"role": "user", "content": "Hello"}]
  }'
```

Streaming version:

```bash
curl https://sccclm.com/en/v1/chat/completions \
  -H "Authorization: Bearer $NEWCCLM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "newcclm-fast",
    "stream": true,
"messages": [{"role": "user", "content": "Hello"}]
  }'
```

It is normal to return a JSON containing `choices`.

---

## 4. DeepSeek Harness (DSH) configuration

Add a custom provider in Settings → Models. **Do not use DeepSeek official or Anthropic preset directories. **

```yaml
# ~/.dsh/settings.yaml
llm-pi-ai:
  providers:
    newcclm:
      displayName: Newcclm
      api: openai-completions
      baseURL: https://sccclm.com/en/v1
      apiKeyEnv: NEWCCLM_API_KEY
      compat:
        supportsDeveloperRole: false
        maxTokensField: max_tokens
      models:
        - id: newcclm-fast
        - id: newcclm-coding
```

```bash
export NEWCCLM_API_KEY="YOUR_API_KEY"
```

Common mistakes in writing:

- `api: anthropic-messages` - Wrong protocol, unable to connect
- `baseURL: https://sccclm.com/en/v1/chat/completions` - repeated path
- fill in real model names such as `deepseek-chat` in models - it will report that the model does not exist. When

---

# # 5. Command Line Client

```bash
npx newcclm # Enter the conversation
npx newcclm -p "question" # Exit after a single question
npx newcclm --status # Check account and remaining usage
npx newcclm --register # Register account
npx newcclm --login # Switch account
npx newcclm --logout # Log out
npx newcclm --help # View usage
```

Available commands in the dialog: `/fast`, `/smart`, `/clear`, `/quit`.

Login credentials are saved in `~/.newcclm/config.json` (readable only by me).
This client does not require an API Key.

---

## 6. Billing A string of characters starting with

| Item | Description |
| --- | --- |
| Profile | ¥30 / 30 days, or $5 / 30 days |
| Monthly quota | $4 usage |
| The quota is exhausted | It will not be interrupted and can be replenished to continue |
| Auto Renew | None |
| Small team profile | ¥300 / starting from 30 days, multiple accounts, invoices can be issued |
| Query usage | `npx newcclm --status` or https://sccclm.com/en/account/ |

---

# # 7. Error Comparison Table

| Return / Phenomenon | Meaning | Recommended processing |
| --- | --- | --- |
| `Invalid token` | Key is wrong, missing or expired | Check key integrity; client user logs in again |
| `model not found` | The model name is not the file name of this service | You can only fill in `newcclm-fast` / `newcclm-coding` |
| `404 Not Found` | Wrong Base URL | Should be `https://sccclm.com/en/v1`, do not superimpose `/v1` |
| The message contains `quota` / `balance` | The current month's usage has been exhausted | Replenish the usage, or wait for the next billing cycle |
| `429 Too Many Requests` | Trigger current limiting | Reduce concurrency and try again later |
| `401` and the message contains `upstream` | Server upstream failure, **non-user configuration problem** | Try again later; do not modify your configuration repeatedly |
| `both_paths_failed` | Both primary and backup lines are unavailable (server failure) | Try again later |
| Connection timeout | Network problem | Check the network environment; if only this service is unreachable, contact the service provider |

**Note**: This service distinguishes between "user configuration problems" and "server failures". If the error message appears
The words `upstream` indicate that the problem is on the server side and the user does not need to modify the configuration.

---

# # 8. Relevant address

- Home page: https://sccclm.com/en/
- Help Center: https://sccclm.com/en/help/
- Client Guide: https://sccclm.com/en/help/client.html
- DSH configuration: https://sccclm.com/en/help/dsh.html
- FAQ: https://sccclm.com/en/help/faq.html
- Check usage: https://sccclm.com/en/account/
- Concise version: https://sccclm.com/en/llms.txt

---

# # 9. Disclaimer

Newcclm is an independent third-party tool service that is not affiliated with the vendor of the model being invoked.
The names and trademarks related to the model belong to their respective owners.