Private Pipeline

The private pipeline is the request boundary used by the recommendation pages. It keeps raw node reads local and sends a reduced payload for scoring.

Where it is used

• Channel Fees
• Opening Recommendations

Flow

1. collect node data through LNC
2. normalize it into a stable internal shape
3. apply the privacy transform used for scoring
4. let the operator review the outgoing request
5. send the reduced payload through Local or Verified Runtime (TEE)
6. show the result and, in verified mode, the verification details

Privacy mode used for scoring

The main scoring mode is feature_only.

That mode keeps the signals needed for scoring, such as:

• references like channel_0001 and peer_0001
• liquidity ratios
• activity and revenue signals
• mission reliability
• centrality and peer aggregates

It avoids sending the full raw node state when that detail is not needed.

Local and verified

Both execution modes use the same local preparation flow. The difference is where the reduced payload is sent:

• Local: the local API
• Verified Runtime (TEE): the verified service path

The internal contracts behind that flow are summarized in Architecture.

Next source-proof layer

The next verified-source step is a collection receipt that binds:

• the exact RPC set
• the raw snapshot hash
• the normalized snapshot hash
• the privacy-transformed snapshot hash

That work is summarized in Source Collection Receipts.