> ## Documentation Index
> Fetch the complete documentation index at: https://docs.blindference.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# Configuration

> All configuration options and environment variables

# Configuration

All configuration is stored in `~/.blindference/config.json` and can be overridden via environment variables prefixed with `BLF_`.

## Default Configuration

```json theme={null}
{
  "node_address": "0x...",
  "keystore_path": "~/.blindference/keystore.json",
  "tier": 0,
  "supported_model_ids": ["facebook/opt-125m"],
  "custom_backends": [],
  "attestation_backend": "mock",
  "icl_endpoint": "https://icl.blindference.xyz",
  "fhenix_rpc": "https://testnet.fhenix.zone",
  "ipfs_gateway": "https://node.lighthouse.storage",
  "model_cache_dir": "~/.blindference/models",
  "log_level": "INFO",
  "network": "fhenix_testnet",
  "attestation_cert_hash": "",
  "attestation_expiry": 0,
  "stake_amount_wei": 0,
  "cofhe_mode": "bridge",
  "cofhe_endpoint": "https://arb-sepolia.g.alchemy.com/v2/demo",
  "cofhe_chain_id": 421614,
  "skip_output_key_storage": false
}
```

## Environment Variables

Every config field can be overridden via an environment variable. The mapping is: `BLF_` + uppercase field name.

| Variable                  | Type   | Default                                     | Description                                         |
| ------------------------- | ------ | ------------------------------------------- | --------------------------------------------------- |
| `BLF_NODE_ADDRESS`        | string | auto-generated                              | Ethereum address (0x...)                            |
| `BLF_KEYSTORE_PATH`       | string | `~/.blindference/keystore.json`             | Path to encrypted keystore                          |
| `BLF_TIER`                | int    | 0                                           | Attestation tier (0=mock, 1=TPM, 2=TEE)             |
| `BLF_SUPPORTED_MODEL_IDS` | list   | `facebook/opt-125m`                         | Comma-separated model IDs                           |
| `BLF_CUSTOM_BACKENDS`     | list   | (empty)                                     | Dotted Python paths for custom backends             |
| `BLF_ATTESTATION_BACKEND` | string | `mock`                                      | `mock`, `tpm`, or `sgx`                             |
| `BLF_ICL_ENDPOINT`        | string | `https://icl.blindference.xyz`              | ICL base URL                                        |
| `BLF_FHENIX_RPC`          | string | `https://testnet.fhenix.zone`               | EVM RPC endpoint                                    |
| `BLF_IPFS_GATEWAY`        | string | `https://node.lighthouse.storage`           | IPFS download/upload gateway                        |
| `BLF_LOG_LEVEL`           | string | `INFO`                                      | `DEBUG`, `INFO`, `WARNING`, `ERROR`                 |
| `BLF_COFHE_MODE`          | string | `bridge`                                    | `bridge` (TypeScript subprocess) or `python` (HTTP) |
| `BLF_COFHE_ENDPOINT`      | string | `https://arb-sepolia.g.alchemy.com/v2/demo` | CoFHE/EVM RPC endpoint                              |
| `BLF_COFHE_CHAIN_ID`      | int    | 421614                                      | Chain ID for CoFHE (Arbitrum Sepolia)               |
| `BLF_KEY_PASSWORD`        | string | (prompted)                                  | Keystore decryption password                        |

## CoFHE Modes

### Bridge Mode (Default)

Spawns a TypeScript subprocess via `@cofhe/sdk/node` for CoFHE operations.

* More reliable SDK lifecycle
* Handles WASM loading correctly
* Better error handling
* Slightly higher overhead

```bash theme={null}
BLF_COFHE_MODE=bridge blindference-node run
```

### Python Mode (Alternative)

Direct HTTP calls to CoFHE endpoints.

* Lighter weight
* Requires manual session management
* Good for resource-constrained environments

```bash theme={null}
BLF_COFHE_MODE=python blindference-node run
```

## Model IDs

Supported models for inference:

| Model ID                       | Provider | Type | Notes                     |
| ------------------------------ | -------- | ---- | ------------------------- |
| `facebook/opt-125m`            | Local    | vLLM | \~125M params, 0.5GB VRAM |
| `groq:llama-3.3-70b-versatile` | Groq     | API  | Fast, requires API key    |
| `gemini:gemini-2.5-flash`      | Google   | API  | Requires API key          |

Set multiple models:

```bash theme={null}
BLF_SUPPORTED_MODEL_IDS=qwen2.5-7b,groq:llama-3.3-70b-versatile,gemini:gemini-2.5-flash
```

## Example: Full Non-Interactive Setup

```bash theme={null}
export BLF_PRIVATE_KEY=0x...
export BLF_KEY_PASSWORD=secure_password
export BLF_ICL_ENDPOINT=https://icl.blindference.xyz
export BLF_COFHE_ENDPOINT=https://arb-sepolia.g.alchemy.com/v2/YOUR_KEY
export BLF_SUPPORTED_MODEL_IDS=facebook/opt-125m,groq:llama-3.3-70b-versatile

blindference-node init
blindference-node run
```

## Example: Local Development Stack

```bash theme={null}
# ICL running locally
export BLF_ICL_ENDPOINT=http://127.0.0.1:9000

# Custom IPFS gateway
export BLF_IPFS_GATEWAY=https://gateway.pinata.cloud/ipfs

# Debug logging
export BLF_LOG_LEVEL=DEBUG

blindference-node run
```