nips/90.md
2023-07-25 17:31:31 +03:00

15 KiB

NIP-90

Data Vending Machine

draft optional author:pablof7z

This NIP defines the interaction between customers and Service Providers for performing on-demand computation.

Money in, data out.

Kinds

This NIP reserves the range 65000-66000 for data vending machine use.

Kind Description
65000 Job feedback
65001 Job result
65002-66000 Job request kinds

Appendix 2 defines the job request types.

Rationale

Nostr can act as a marketplace for data processing, where users request jobs to be processed in certain ways (e.g., "speech-to-text", "summarization", etc.), but they don't necessarily care about "who" processes the data.

This NIP is not to be confused with a 1:1 marketplace; instead, it describes a flow where a user announces a desired output, willingness to pay, and service providers compete to fulfill the job requirement in the best way possible.

Actors

There are two actors in the workflow described in this NIP:

  • Customers (npubs who request a job)
  • Service providers (npubs who fulfill jobs)

Event Kinds

  • kind:65000: job feedback
  • kind:65001: job result
  • kind:65002-kind:66000: job requests

Job request

A request to have data processed, published by a customer. This event signals that an npub is interested in receiving the result of some kind of compute.

{
    "kind": 65xxx, // kind in 65002-66000 range
    "content": "",
    "tags": [
        [ "i", "<data>", "<input-type>", "<marker>", "<relay>" ],
        [ "output", "<mime-type>" ],
        [ "relays", "wss://..."],
        [ "bid", "<msat-amount>" ],
        [ "t", "bitcoin" ]
    ]
}

All tags are optional.

  • i tag: Input data for the job (zero or more inputs)
    • <data>: The argument for the input
    • <input-type>: The way this argument should be interpreted. MUST be one of:
      • url: A URL to be fetched
      • event: A Nostr event ID.
      • job: The output of a previous job with the specified event ID
      • text: <data> is the value of the input, no resolution is needed
    • <marker>: An optional field indicating how this input should be used within the context of the job
    • <relay>: If event or job input-type, the relay where the event/job was published, otherwise optional or empty string
  • output: Expected output format. (e.g. MIME type)
    • Service Providers MUST publish the result of the job in this format if it has been specified.
    • Each job-type (Appendix 2) might define the output format more narrowly.
  • bid: Customer MAY specify a maximum amount (in millisats) they are willing to pay
  • relays: List of relays where Service Providers SHOULD publish responses to
  • p: Service Providers the customer is interested in. Other SPs MIGHT still choose to process the job

Job result

Service providers publish job results, providing the output of the job result. They should tag the original job request event id as well as the customer's pubkey.

{
    "pubkey": "<service-provider pubkey>",
    "content": "<payload>",
    "kind": 65001,
    "tags": [
        [ "request", "<job-request>" ],
        [ "e", "<job-request-id>", "<relay-hint>" ],
        [ "p", "<customer's-pubkey>" ],
        [ "amount", "requested-payment-amount", "<optional-bolt11>" ]
    ]
}
  • request: The job request event stringified-JSON.
  • amount: millisats that the Service Provider is requesting to be paid. An optional third value can be a bolt11 invoice.

Job feedback

Service providers can give feedback about a job back to the customer.

{
    "kind": 65000,
    "content": "<empty-or-payload>",
    "tags": [
        [ "status", "<status>", "<extra-info>" ],
        [ "amount", "requested-payment-amount", "<bolt11>" ],
        [ "e", "<job-request-id>", "<relay-hint>" ],
        [ "p", "<customer's-pubkey>" ],
    ]
}
  • content: Either empty or a job-result (e.g. for partial-result samples)
  • status tag: Service Providers SHOULD indicate what this feedback status refers to. Appendix 3 defines status. Extra human-readable information can be added as an extra argument.
  • amount tag: as defined in the Job Result section.

Protocol Flow

  • Customer publishes a job request (e.g. kind:65002 speech-to-text).
  • Service Providers can submit kind:65000 job-feedback events (e.g. payment-required, processing, error, etc.).
  • Upon completion, the service provider publishes the result of the job with a kind:65001 job-result event.
  • At any point, if there is an amount pending to be paid as instructed by the service provider, the user can pay the included bolt11 or zap the job result event the service provider has sent to the user

kind:65000 and kind:65001 events MAY include an amount tag, this can be interpreted as a suggestion to pay. Service Providers SHOULD use the payment-required feedback event to signal that a payment is required and no further actions will be performed until the payment is sent. Customers can always either pay the included bolt11 invoice or zap the event requesting the payment and service providers should monitor for both if they choose to include a bolt11 invoice.

Notes about the protocol flow

The flow is deliberately ambiguous, allowing vast flexibility for the interaction between customers and service providers so that service providers can model their behavior based on their own decisions/perceptions of risk.

Some service providers might choose to submit a payment-required as the first reaction before sending a processing or before delivering kind:65001 results, some might choose to serve partial results for the job (e.g. as a sample), send a payment-required to deliver the rest of the results, and some service providers might choose to assess likelihood of payment based on an npub's past behavior and thus serve the job results before requesting payment for the best possible UX.

It's not up to this NIP to define how individual vending machines should choose to run their business.

Cancellation

A job request might be cancelled by publishing a kind:5 delete request event tagging the job request event.

Job chaining

A Customer MAY request multiple jobs to be processed as a chain, where the output of a job can be the input of another job. (e.g. podcast transcription -> summarization of the transcription). This is done by specifying as input an event id of a different job with the job type.

Service Providers MAY begin processing a subsequent job the moment they see the prior job's result, but they will likely wait for a zap to be published first. This introduces a risk that Service Provider of job #1 might delay publishing the zap event in order to have an advantage. This risk is up to Service Providers to mitigate or to decide whether the service provider of job #1 tends to have good-enough results so as to not wait for an explicit zap to assume the job was accepted.

This gives a higher level of flexibility to service providers (which sophisticated service providers would take anyway).

Consult Appendix 1: Example's Summarization of a podcast

E.g. Payment required (with sample content)

{
    "kind": 65000,
    "content": "This is the transcription service that you",
    "tags": [
        [ "e", <job-request-id>, <relay-hint> ],
        [ "p", <customer-pubkey> ],
        [ "status", "payment-required" ],
        [ "amount", "7000" ],
    ]
}

Appendix 1: Examples

Transcript of a podcast from second 900 to 930.

kind:65002: Speech-to-text job request

{
    "id": "12345",
    "pubkey": "abcdef",
    "content": "",
    "kind": 65002,
    "tags": [
        [ "i", "https://bitcoin.review/episode1.mp3", "url" ],
        [ "params", "range", "900", "930" ],
        [ "output", "text/plain" ],
        [ "bid", "50000" ],
    ]
}

kind:65000: Job Feedback: request for (partial) payment

  • The SP is signaling here that it won't start processing until 100 sats are paid
{
    "kind": 65000,
    "content": "",
    "tags": [
        ["e", "12345"],
        ["p", "abcdef"],
        ["status", "payment-required"],
        ["amount", "100000"]
    ]
}
  • User zaps 100 sats to the kind:65000 job-feedback

kind:65001: Job result + request for remaining payment

{
    "content": "blah blah blah",
    "tags": [
        ["e", "12345"],
        ["p", "abcdef"],
        ["amount", "400000"]
    ]
}

Summarization of a podcast

User publishes two job requests at the same time. A job that transcribes an audio and a job that summarizes the transcription (output of job #1).

User publishes event #1 and #2 together.

kind:65002: Job Request #1: speech-to-text

{
    "id": "12345",
    "pubkey": "abcdef",
    "kind": 65002,
    "content": "",
    "tags": [
        [ "i", "https://bitcoin.review/episode1.mp3", "url" ],
        [ "output", "text/plain" ],
        [ "params", "range", "900", "930" ],
        [ "bid", "100000" ]
    ]
}

kind:65002: Job Request #2: summarization of job #1's result

{
    "id": "12346",
    "pubkey": "abcdef",
    "kind": 65003,
    "content": "",
    "tags": [
        [ "i", "12345", "job" ], // input is the output of job with id 12345
        [ "output", "text/plain" ],
        [ "params", "length", "3 paragraphs" ],
        [ "bid", "10000" ]
    ]
}

Translation of a note

kind:65004: Job Request #1: translation of an existing note

{
    "id": "12346",
    "pubkey": "abcdef",
    "content": "",
    "kind": 65004,
    "tags": [
        [ "i", "<hexid>", "event", "wss://relay.nostr.com" ]
        [ "output", "text/plain" ],
        [ "params", "lang", "es_AR" ],
        [ "bid", "5000" ]
    ]
}

kind:65001: Job result

{
    "kind": 65001,
    "content": "Che, que copado, boludo!",
    "tags": [
        ["e", "12346"],
        ["p", "abcdef"],
        ["amount", "4000"]
    ]
}

AI-image of the summarization of 2 podcasts

kind:65002: Job request #1 (transcribe podcast #1)

{
    "id": "123",
    "kind": 65002,
    "tags": [
        [ "i", "https://example.com/episode1.mp3", "url" ],
        [ "bid", "100000" ]
    ]
}

kind:65002: Job request #2 (transcribe podcast #2)

{
    "id": "124",
    "kind": 65002,
    "tags": [
        [ "i", "https://example.com/episode2.mp3", "url" ],
        [ "bid", "100000" ]
    ]
}

kind:65003: Job request #3 (summarize the two job's outputs into one paragraph)

{
    "id": "125",
    "kind": 65003,
    "tags": [
        [ "param", "length", "1 paragraph" ],
        [ "i", "123", "job" ],
        [ "i", "124", "job" ],
        [ "bid", "100000" ]
    ]
}

kind:65005: Job request #4 (generate image based on the summary)

{
    "id": "126",
    "kind": 65004,
    "tags": [
        [ "i", "125", "job" ],
        [ "param", "prompt", "photorealistic" ],
        [ "param", "size", "4000x4000" ],
        [ "bid", "500000" ]
    ]
}

AI-image of embedded input

kind:65005: Job request

{
    "kind": 65004,
    "tags": [
        [ "i", "Millions of vending machines, interconnected with tubes with eah other", "text" ],
        [ "param", "prompt", "photorealistic" ],
        [ "bid", "500000" ]
    ]
}

Appendix 2: Job types

This is a list of all the supported job requests.

speech-to-text: kind:65002

params

param req? description
range opt timestamp range (in seconds) of desired text to be transcribed
alignment opt word, segment, raw: word-level, segment-level, or raw outputs

summarization: kind:65003

param req? description
length opt desired length

translation: kind:65004

param req? description
lang req desired language in BCP 47 format.

image generation: kind:65005

param req? description
prompt opt extra prompt to be used for the image generation
size opt desired size of the image

event list generation: kind:65006

Generates a list of event ids, (e.g. algorithmic feeds, spam-free notifications, trending topics)

Output should be a stringified array of elements usually find in a nostr event' tags, e.g.:

{ "content": "[
    [\"e\", \"<id>\"],
    [\"a\", \"30023:pubkey:id\"],
    [\"t\", \"tag\"],
    [\"p\", \"pubkey\"],
]" }
param req? description
filter opt Stringified JSON REQ filter. Used to scope the desired results (e.g. specify the desired kinds)
prompt opt A human-readable description of the desired results. Which might be used with e.g. an LLM to tune the results.
p opt Array of pubkeys to generate a feed from someone else's point-of-view. This param allows for a client to choose to generate the feeds and incur the costs of its users.

example job-request

Generate an algorithmic feed of the most interesting kind:1s related to the topic "bitcoin", tagging service providers specializing in safe-for-work content that would interest pubkey pubkey1.

{
    "kind": 65006,
    "tags": [
        [ "param", "filter", "{ \"kinds\": [1], \"#t\": [\"bitcoin\"] }" ],
        [ "param", "p", "[\"pubkey1\"]"]
        [ "bid", "5000" ],
        [ "t", "sfw" ]
    ]
}

Appendix 3: Job feedback status

status description
payment-required Service Provider requires payment before continuing.
processing Service Provider is processing the job.
error Service Provider was unable to process the job.
success Service Provider successfully processed the job.
partial Service Provider partially processed the job. The .content might include a sample of the partial results.

Any job feedback event MIGHT include results in the .content field, as described in the Job Result section. This is useful for service providers to provide a sample of the results that have been processed so far.

Appendix 4: Service provider discoverability

Service Providers can use NIP-89 announcements to advertise their support for job kinds:

{
    "kind": 31990,
    "pubkey": "<pubkey>",
    "tags": [
        [ "k", 65002 ], // e.g. speech-to-text
        [ "t", "bitcoin" ] // e.g. optionally advertises it specializes in bitcoin audio transcription that won't confuse "Drivechains" with "Ridechains"
    ]
}

Customers can use NIP-89 to see what service providers their follows use.