Skip to content
TAO$188.32
BTC0.002928
MCap$2.22B
Vol$86.25M

API reference / Subnets

One subnet's GitHub repository on bittensor.ai: stars, activity, contributors, releases, analysis

GET/api/v2/subnets/{netuid}/repo
Status
Beta
Auth
API key or OAuth access token
Scope
data:read
Rate limit
plan

The GitHub repository served for the subnet that holds netuid now, with its figures.

Full description

The GitHub repository served for the subnet that holds netuid now, with its figures (stars, forks, open issues, watchers, 4-week commits, contributors, language, licence, default branch, the last push, whether GitHub still serves it) as the off-chain source last observed them on GitHub, up to 30 contributors, the 5 newest releases (metadata only) and bittensor.ai's newest automated analysis (activity status, documentation score, release cadence, contributor concentration; freshness figures re-based to now). No chain figures: the chain side of the page is GET /api/v2/subnets/{netuid}. Which repository (repo.repo_source): the one the subnet's owner set on their bittensor.ai page for this life (owner_claim), else the one the life's on-chain identity names (chain_identity), else bittensor.ai's registry's (registry); a claimed or identity value that names no real repository (malformed, another host, an org page, a placeholder, or gone from GitHub while the next source's is live) is passed over and repo.repo_source_detail says why. Figures offchain cannot state are null with repo.figures_reason (not_mirrored, not_observed, stale, upstream_gone, offchain_unavailable), never an older value. Current life by default: the registry's repository is matched to the head directory row's generation by the rule include=meta uses for meta.repo_url (stamped with that life, or unstamped while the slot's record is this life's or the live on-chain identity names it), so a re-registered slot never shows the previous subnet's repository: with no claim or identity, repo is then null with reason (no_repository, previous_life, unattributed), a 200, and previous_generation names the life it belongs to. The life comes from the same head answer as GET /api/v2/subnets/names (no extra data-source call when the directory was served in the last 5 minutes). Life attribution is by time: a repository, sync or analysis belongs to the life whose window on GET /api/v2/subnets/generations (start_time to end_time, read once for every slot and kept 10 minutes) holds its link, sync or analysis time, whatever life it was stamped with when written; generation_stamp, analysis.generation, details.generation and scans[].generation then name that life, and previous_generation the canonical id of the life holding a withheld repository's or analysis' newest row. generation (a life's id or its start height) serves one life of the slot instead (an ended subnet's own page): the life is checked against GET /api/v2/subnets/generations (one data-source read) and the answer adds life_start_height, life_start_time, life_end_height, life_end_time and ended. For an ended life everything is as of that life, never the current one: its owner's claim, its last on-chain identity's repository, else the registry's linked for it (stamped with it, or named by its archived record), offchain's newest observation before its end, the releases published before its end, the newest analysis written for it (analysis_reason no_analysis_for_generation when none; freshness figures re-based to its end); contributors are offchain's newest list at or before its end. include=details adds data.details, the same analysis' content (miner and validator hardware with each value's quoted source, setup steps and their review in setup_review (each step's source line, file checks and risk; dangerous commands withheld), incentive mechanism, architecture, entry points, config parameters, dependencies, the Markdown sections, SEO title / description / keywords / problem and solution / use cases / FAQs), under the same life rule (null with repo). include=scans adds data.scans, the life's analysis runs (up to 50, newest first, no content). Cached 5 minutes (slow class; offchain observes GitHub every few hours), an ended life's answer 1 hour. Errors are application/problem+json: INVALID_NETUID, SUBNET_NOT_FOUND (404: no subnet is registered in the slot at the head; with generation, the slot was never registered), INVALID_GENERATION (400), GENERATION_NOT_FOUND (404: the slot never had that life), INVALID_INCLUDE, UNKNOWN_PARAMETER, NOT_READY (503, generation only: the lives registry does not reach genesis yet), SOURCE_TIMEOUT (503), SOURCE_NOT_READY (503).

Beta: this route may change with one release's notice in the v2 changelog; additive changes need none (x-stability: beta).

Path parameters

netuidpath, required· integer · 0 to 65535

Subnet slot (0 = root).

Query parameters

generationquery· integer · at least 0

The subnet life to serve: its generation as GET /api/v2/subnets/generations and the subnet rows name it (the stored registration id), or its start_height (so every previous_generation resolves); the id is matched first, and data.generation is always the life's canonical id. Omitted: the current life, the answer unchanged. Neither of a life of the slot: a 404 GENERATION_NOT_FOUND; not an integer >= 0: a 400 INVALID_GENERATION.

includequery· string

A comma list (repeatable): details, the analysis' content in data.details; scans, the life's analysis runs in data.scans (each absent without it). Anything else is a 400 INVALID_INCLUDE.

Response

data is an object with these fields. Amounts are integer strings in their base unit, and null means unknown, never zero (see Units and formats). meta.at names the block the answer was read at.

The slot's subnet life and its repository.

netuidinteger · unit id

Subnet slot (0 = root).

generationinteger · unit id · can be null

Opaque life id (the life's registration identifier), NOT a block height; use life_start_height for the first block. The subnet life the repository was matched to: the head directory row's generation (as on GET /api/v2/subnets/{netuid}), or the life the generation parameter asked for, always its canonical id as on GET /api/v2/subnets/generations (also when asked by its start height). Null when the data source has not recorded it.

is_current_lifeboolean · can be null

Whether generation is the slot's current life (true) or an ended one, re-registered since (false). Null exactly when generation is null, or transiently when this API cannot read the lives. Root's generation 0 is current.

reasonstring · can be null

Why repo is null; null when it is served. Only when neither the owner's claim nor the on-chain identity names a repository for the life (else that one is served), the registry's answer: no_repository: bittensor.ai tracks no repository for the slot (for an ended life: none linked for that life). previous_life (current life only): what it tracks belongs to an earlier subnet in this slot (the slot's record, or every linked repository, is stamped with another registration; see previous_generation); it is never served for this life. unattributed: a repository is linked but nothing shows it is this life's (not stamped, the slot's record is not this life's and the on-chain identity does not name it; for an ended life, neither its archived record nor its last identity names it). New values may be added; treat unknown values as unavailable.

repoobject · can be null

The repository served for this life: its owner's claim, else its on-chain identity's, else the registry's newest one attributable to it (repo.repo_source). Its figures are the off-chain source's; for an ended life the newest observation before its end (null with repo.figures_reason when none). Null with reason (no source names a repository for this life).

22 fields in repo
ownerstring

The GitHub owner (user or organisation): as our registry spells it when it tracks the repository, else lower-cased (GitHub treats owner and name case-insensitively).

namestring

The repository name, spelled as owner.

urlstring

https://github.com/{owner}/{name}.

github_idinteger · unit id · can be null

GitHub's numeric repository id (stable across renames): bittensor.ai's registry's when it tracks the repository, else the off-chain source's. Null only when neither has fetched it yet.

generation_stampinteger · unit id · can be null

The subnet life the repository was linked for (its registration block, as generation). For repo_source owner_claim / chain_identity it is that life (the claim and the identity are the life's own). For registry: the link's stamp, which then equals data.generation; null for a repository linked before stamping, served because the slot's record is this life's or the live on-chain identity names it. When the life's time window is known the registry's repository is attributed by time, and a stamped repository's member is that life (a stored stamp that the window contradicts is never served).

repo_sourcestring

Which source chose the repository, in precedence order: owner_claim (the repository the subnet's owner set on their bittensor.ai page for this life), chain_identity (the repository the life's on-chain identity names: the live identity, or an ended life's last one), registry (bittensor.ai's curated registry, attributed to the life by its generation guard). New values may be added; show an unknown value as is (or generically).

repo_source_detailstring · can be null

Why a higher-precedence source was passed over, <source>_rejected: <why> joined by ; (e.g. chain_identity_rejected: org_page). <why>: malformed (names no repository), not_github (another host), org_page (a GitHub site page such as github.com/orgs/<org>), placeholder (a stand-in such as deprecated/deprecated), not_found (GitHub no longer serves it while the next source's repository is live). Null when nothing was passed over.

upstream_statestring · can be null

Whether GitHub still serves the repository, as the off-chain source last checked: ok, not_found (404, private or taken down) or archived. Null when offchain has not checked it (or with figures_reason not_mirrored / offchain_unavailable). New values may be added; show an unknown value as is (or generically).

descriptionstring · can be null

GitHub's repository description, from bittensor.ai's registry (only when it tracks this repository). Null when none.

languagestring · can be null

GitHub's primary language. Null with figures_reason, or when GitHub states none.

licensestring · can be null

The licence GitHub detected (SPDX id; NOASSERTION for a non-SPDX licence). Null with figures_reason, or when GitHub states none.

topicsarray of string

GitHub's topics, from bittensor.ai's registry ([] when none, not synced yet, or the registry does not track this repository).

default_branchstring · can be null

The default branch. Null with figures_reason, or when GitHub states none.

starsinteger · unit count · can be null

Stargazers. Null with figures_reason, or when GitHub states none.

forksinteger · unit count · can be null

Forks. Null with figures_reason, or when GitHub states none.

open_issuesinteger · unit count · can be null

Open issues and pull requests (GitHub counts both). Null with figures_reason, or when GitHub states none.

watchersinteger · unit count · can be null

Watchers (subscribers). Null with figures_reason, or when GitHub states none.

commit_count_30dinteger · unit count · can be null

Commits on the default branch in the 4 weeks before last_synced_at (GitHub's weekly participation counts; null until GitHub has computed them). Null with figures_reason, or when GitHub states none.

contributor_countinteger · unit count · can be null

Contributors GitHub lists for the repository (its first 100, anonymous excluded), as of the contributor list served in contributors. Null with figures_reason, or when GitHub states none.

pushed_atstring (date-time) · can be null

GitHub's pushed_at: the latest push to any branch. Null with figures_reason, or when GitHub states none.

Format: date and time, ISO 8601 in UTC
last_synced_atstring (date-time) · can be null

When the off-chain source observed the figures above on GitHub (for an ended life: the newest observation before its end). Null with figures_reason.

Format: date and time, ISO 8601 in UTC
figures_reasonstring · can be null

Why the figures (upstream_state excepted when known) are null; null when they are served. not_mirrored: bittensor.ai's off-chain source does not track the repository yet. not_observed: it tracks it but observed no figures for the period (an ended life before its repository was first observed: history runs from 2026-04-15 for most repositories, weekly before July, and from their first observation otherwise). stale: its newest observation is over 48 hours old, or its data has not synced for 6 hours. upstream_gone: GitHub no longer serves the repository. offchain_unavailable: the off-chain data source could not be read here. Never an older value from another source. New values may be added; treat unknown values as unavailable.

contributorsarray of objects

Up to 30 contributors, most contributions first: the off-chain source's newest contributor list of the repository (for an ended life, the newest at or before its end). [] without repo, or with repo.figures_reason when offchain holds no list.

3 fields in contributors (each item)
loginstring

The contributor's GitHub login.

avatar_urlstring · can be null

Their GitHub avatar image URL. Null when none.

contributionsinteger · unit count

Their commits on the default branch, as GitHub counts them.

releasesarray of objects

The 5 newest releases, newest first (pre-releases included; releases deleted upstream left out), from the off-chain source; for an ended life, the newest published before its end. [] without repo, releases, or when offchain could not be read.

7 fields in releases (each item)
tag_namestring

The release's git tag.

namestring · can be null

The release title. Null when none, or when is_redacted.

bodystring · can be null

The release notes as published (GitHub Markdown, raw and untrusted: sanitise when rendering). Null when none, or when is_redacted.

published_atstring (date-time)

When the release was published (its creation time, or when it was first seen, when GitHub states none).

Format: date and time, ISO 8601 in UTC
is_prereleaseboolean

Whether GitHub marks it a pre-release.

urlstring

The release page on GitHub (.../releases/tag/{tag_name}).

is_redactedboolean

True when its author asked for their personal data to be removed: the release is kept (counts stay right) with its tag and time, without its title and notes.

analysisobject · can be null

The newest analysis of the repository written for this life: stamped with generation, or written before stamping while the repository is stamped with it. For an ended life its freshness figures are re-based to the life's end, not now. Null without repo (see reason) or with analysis_reason.

12 fields in analysis
generationinteger · unit id · can be null

Opaque life id (the life's registration identifier), NOT a block height; use life_start_height for the first block. The subnet life the analysis was written for (equal to data.generation when stamped). Null for an analysis written before stamping, served because the repository itself is stamped with this life. When the life's time window is known the analysis is attributed by its time, and this is that life (a stored stamp that the window contradicts is never served).

life_start_heightinteger · unit block · can be null

The first block of the life generation names (GET /subnets/generations' start_height): use it, never generation, for heights, ages and dates. Null exactly when generation is null (root without a life, no subnet), or transiently when this API cannot read the lives. Root's generation 0 starts at 0.

is_current_lifeboolean · can be null

Whether generation is the slot's current life (true) or an ended one, re-registered since (false). Null exactly when generation is null, or transiently when this API cannot read the lives. Root's generation 0 is current.

statusstring

How the analysis run ended. Known values: complete, partial, failed. New values may be added; show an unknown value as is (or generically).

analyzed_atstring (date-time)

When the analysis ran.

Format: date and time, ISO 8601 in UTC
summarystring · can be null

The analysis' description of the subnet. Null when none.

activity_statusstring

From commit_staleness_days: active under 30 days, stale under 180, abandoned otherwise (also when unknown). New values may be added; show an unknown value as is (or generically).

commit_staleness_daysinteger · unit days · can be null

Whole days from repo.pushed_at to now (the analysis' own figure when the push time is unknown). Null when neither is known.

doc_quality_scoreinteger · unit score

Documentation quality, 0..100, scored by the analysis.

release_frequency_daysstring · unit days · 1 decimals · can be null

Mean days between the releases seen by the analysis, 1 decimal. Null with fewer than two.

Format: decimal number, as a string
days_since_last_releaseinteger · unit days · can be null

Whole days from the newest release to now (the analysis' own figure when no release is stored). Null without a release.

contributor_concentration_ratiostring · unit ratio · 2 decimals · can be null

The top contributor's share of all contributions, 0..1, 2 decimals. Null when unknown.

Format: decimal number, as a string
analysis_reasonstring · can be null

Why analysis (and details) is null while repo is served; null otherwise. analysis_previous_life: the repository's analyses were all written for an earlier subnet in this slot (a project that kept its repository across a re-registration; see previous_generation); they are never served for this life. analysis_pending: no analysis has been written for this life yet (or the data source has not recorded generation, or the served repository is not one bittensor.ai's registry tracks). no_analysis_for_generation (an ended life, generation given): no analysis was written during that life. analysis_unavailable: the analysis source could not be read this time; retry shortly (scans is then []). New values may be added; treat unknown values as unavailable.

previous_generationinteger · unit id · can be null

With reason previous_life or analysis_reason analysis_previous_life: the life the withheld repository or analysis belongs to: the canonical generation of the life whose time window holds its newest row, else its stamp as stored (that life's registration block, usually its generation; a stamp from the lifecycle registry is its start_height). Either form is accepted by this route's generation parameter, whose answer names the canonical id in data.generation, so the page can link that life's own page. Null otherwise.

life_start_heightinteger · unit block · can be null

Only with the generation parameter (absent otherwise): the life's first block (start_height of GET /api/v2/subnets/generations).

life_start_timestring (date-time) · can be null

Only with generation: the time of life_start_height, RFC 3339 UTC. Null when not known.

Format: date and time, ISO 8601 in UTC
life_end_heightinteger · unit block · can be null

Only with generation: the block that ended the life, exclusive (end_height of GET /api/v2/subnets/generations). Null while it is the current life.

life_end_timestring (date-time) · can be null

Only with generation: the time of life_end_height, RFC 3339 UTC. Null while current.

Format: date and time, ISO 8601 in UTC
endedboolean · can be null

Only with generation: true when the life has ended (another subnet holds the slot, or none): everything here is as of that life, to be shown as ended. False for the current life.

detailsobject · can be null

Only with include=details (absent otherwise): the content of the same analysis as analysis (hardware with provenance, setup steps, incentive mechanism, architecture, dependencies, Markdown sections, SEO texts). Null exactly when analysis is, so never another life's write-up.

31 fields in details
generationinteger · unit id · can be null

Opaque life id (the life's registration identifier), NOT a block height; use life_start_height for the first block. The subnet life the analysis was written for (equal to data.generation when stamped). Null for an analysis written before stamping, served because the repository itself is stamped with this life. When the life's time window is known the analysis is attributed by its time, and this is that life (a stored stamp that the window contradicts is never served).

life_start_heightinteger · unit block · can be null

The first block of the life generation names (GET /subnets/generations' start_height): use it, never generation, for heights, ages and dates. Null exactly when generation is null (root without a life, no subnet), or transiently when this API cannot read the lives. Root's generation 0 starts at 0.

is_current_lifeboolean · can be null

Whether generation is the slot's current life (true) or an ended one, re-registered since (false). Null exactly when generation is null, or transiently when this API cannot read the lives. Root's generation 0 is current.

analyzed_atstring (date-time)

When the analysis ran (as analysis.analyzed_at).

Format: date and time, ISO 8601 in UTC
miner_hardwareobject

A miner's hardware requirements, with provenance.

14 fields in miner_hardware
gpu_requiredboolean · can be null

Whether a GPU is required: true / false only when a documentation line says so. Null: unknown.

gpu_modelstring · can be null

The GPU model the documentation names. Null when unknown.

gpu_vram_gbstring · unit gigabytes · 3 decimals · can be null

Minimum GPU memory in GB, a decimal string, at most 3 decimals, trailing zeros dropped. Null when unknown.

Format: decimal number, as a string
gpu_countinteger · unit count · can be null

GPUs required. Null when unknown.

min_ram_gbstring · unit gigabytes · 3 decimals · can be null

Minimum system memory in GB, a decimal string, at most 3 decimals, trailing zeros dropped. Null when unknown.

Format: decimal number, as a string
recommended_ram_gbstring · unit gigabytes · 3 decimals · can be null

Recommended system memory in GB, a decimal string, at most 3 decimals, trailing zeros dropped. Null when unknown.

Format: decimal number, as a string
disk_gbstring · unit gigabytes · 3 decimals · can be null

Minimum disk space in GB, a decimal string, at most 3 decimals, trailing zeros dropped. Null when unknown.

Format: decimal number, as a string
cpu_coresinteger · unit count · can be null

Minimum CPU cores. Null when unknown.

os_requirementsstring · can be null

The operating system requirement (e.g. Ubuntu 22.04). Null when unknown.

network_requirementsstring · can be null

The network requirement (bandwidth, open ports). Null when unknown.

notesstring · can be null

Another stated hardware requirement, as documented. Null when none.

sourcesobject

The source of each served (non-null) field above, keyed by field name.

withheldobject

Fields served null although a value was found or claimed, keyed by field name, with why.

conflictsarray of objects

Fields where min_compute.yml and the documentation disagree, and how each was resolved.

validator_hardwareobject

A validator's hardware requirements, with provenance.

14 fields in validator_hardware
gpu_requiredboolean · can be null

Whether a GPU is required: true / false only when a documentation line says so. Null: unknown.

gpu_modelstring · can be null

The GPU model the documentation names. Null when unknown.

gpu_vram_gbstring · unit gigabytes · 3 decimals · can be null

Minimum GPU memory in GB, a decimal string, at most 3 decimals, trailing zeros dropped. Null when unknown.

Format: decimal number, as a string
gpu_countinteger · unit count · can be null

GPUs required. Null when unknown.

min_ram_gbstring · unit gigabytes · 3 decimals · can be null

Minimum system memory in GB, a decimal string, at most 3 decimals, trailing zeros dropped. Null when unknown.

Format: decimal number, as a string
recommended_ram_gbstring · unit gigabytes · 3 decimals · can be null

Recommended system memory in GB, a decimal string, at most 3 decimals, trailing zeros dropped. Null when unknown.

Format: decimal number, as a string
disk_gbstring · unit gigabytes · 3 decimals · can be null

Minimum disk space in GB, a decimal string, at most 3 decimals, trailing zeros dropped. Null when unknown.

Format: decimal number, as a string
cpu_coresinteger · unit count · can be null

Minimum CPU cores. Null when unknown.

os_requirementsstring · can be null

The operating system requirement (e.g. Ubuntu 22.04). Null when unknown.

network_requirementsstring · can be null

The network requirement (bandwidth, open ports). Null when unknown.

notesstring · can be null

Another stated hardware requirement, as documented. Null when none.

sourcesobject

The source of each served (non-null) field above, keyed by field name.

withheldobject

Fields served null although a value was found or claimed, keyed by field name, with why.

conflictsarray of objects

Fields where min_compute.yml and the documentation disagree, and how each was resolved.

is_staleboolean

True when the analysed repository (or its hardware-docs repository) no longer exists on GitHub (404) or is archived: the values describe the last version seen and may be out of date.

stale_reasonstring · can be null

Why is_stale is true: repo_not_found, repo_archived, hardware_docs_not_found or hardware_docs_archived. null when not stale. Known values: repo_not_found, repo_archived, hardware_docs_not_found, hardware_docs_archived. New values may be added; treat unknown values as unavailable.

miner_setup_instructionsarray of objects

Setting up a miner, step by step.

4 fields in miner_setup_instructions (each item)
stepinteger · unit count

The step's number (1-based, in order).

commandstring · can be null

The shell command to run, as untrusted repository text: show it as plain code, never as a link. Null when the step has no command or the review withheld it (setup_review says why and where it is in the repository).

descriptionstring

What the step does.

optionalboolean

Whether the documentation marks the step optional.

validator_setup_instructionsarray of objects

Setting up a validator, step by step.

4 fields in validator_setup_instructions (each item)
stepinteger · unit count

The step's number (1-based, in order).

commandstring · can be null

The shell command to run, as untrusted repository text: show it as plain code, never as a link. Null when the step has no command or the review withheld it (setup_review says why and where it is in the repository).

descriptionstring

What the step does.

optionalboolean

Whether the documentation marks the step optional.

setup_reviewobject · can be null

The setup steps and entry points, reviewed: where each is stated in the repository, whether the files they name exist, and how risky each is to run (a dangerous command is withheld). Prefer these over miner_setup_instructions, validator_setup_instructions and entry_points. Null with setup_review_reason.

9 fields in setup_review
review_versioninteger · unit id · can be null

The rules version of the review (the vocabularies in it are this version's). Null when not recorded.

risk_levelstring

The riskiest step's level: none, caution or danger. New values may be added; show an unknown value as is (or generically).

flagsarray of string

Every step's risk flags together, sorted. [] when none. Known values: custom_package_index, destructive_command, disables_security, malicious_host, mnemonic_on_command_line, non_https, obfuscated_command, opens_port, other_repo, package_lookalike, pipe_to_shell, pipe_to_shell_unvetted, privileged_container, secret_in_command, spends_tao, sudo, suspicious_host, third_party_package_source, unlisted_package, unvetted_host, wallet_exfiltration, wallet_key_material. New values may be added; show an unknown value as is (or generically).

summaryobject

Counts per role.

miner_stepsarray of objects

Setting up a miner, step by step, reviewed.

validator_stepsarray of objects

Setting up a validator, step by step, reviewed.

entry_pointsarray of objects

How to run each role, reviewed.

requirementsarray of objects

What each role needs before it runs, by name only. [] when none were found.

reviewed_atstring (date-time) · can be null

When the review last changed (its risk is re-checked as hosts are re-vetted). Null when not recorded.

Format: date and time, ISO 8601 in UTC
setup_review_reasonstring · can be null

Why setup_review is null: not_reviewed (this analysis has no review: treat its setup steps as unverified), review_unavailable (the review could not be read this time). Null when served. New values may be added; treat unknown values as unavailable.

incentive_mechanismobject · can be null

How miners are scored and rewarded. Null when the analysis found nothing.

3 fields in incentive_mechanism
summarystring · can be null

A short statement of how miners are scored. Null when none.

detailsstring · can be null

The longer write-up (plain text; the subnet page's incentive section). Null when none.

scoring_dimensionsarray of objects

The scoring criteria with their weights ([] when none are stated).

architecture_summarystring · can be null

How the subnet's components interact (plain text). Null when none.

entry_pointsarray of objects

How to run each role.

3 fields in entry_points (each item)
rolestring

The role (e.g. miner, validator).

commandstring · can be null

The command that runs it, as untrusted repository text: show it as plain code, never as a link. Null when not stated or the review withheld it.

descriptionstring · can be null

What it runs. Null when not stated.

key_config_paramsobject

Configuration parameters users typically change: parameter name to what it controls.

dependenciesarray of objects

The repository's dependencies.

4 fields in dependencies (each item)
namestring

The package name (PEP 503 normalised when parsed from a requirements file).

version_constraintstring · can be null

The version constraint (e.g. >=2.0). Null when unconstrained or unknown.

purposestring · can be null

What it is used for, per the analysis. Null when not stated.

sourcestring · can be null

Where it was found: parser (a requirements / pyproject file), llm (the documentation), both. Null when not recorded. New values may be added; show an unknown value as is (or generically).

python_versionstring · can be null

The Python version required (e.g. >=3.10). Null when unknown.

bittensor_versionstring · can be null

The bittensor SDK version required (e.g. ==9.9.0). Null when unknown.

external_servicesarray of string

External services it depends on (e.g. postgresql, s3, wandb, openai, huggingface, a key-value cache).

md_overviewstring · can be null

Markdown: what the subnet does and the problem it solves. Null when not generated.

md_key_featuresstring · can be null

Markdown: 3-6 bullets of notable features. Null when not generated.

md_how_it_worksstring · can be null

Markdown: the architecture and how miners and validators interact. Null when not generated.

md_incentive_mechanismstring · can be null

Markdown: scoring, rewards and penalties. Null when not generated.

md_deploymentstring · can be null

Markdown: hardware and infrastructure needs. Null when not generated.

seo_titlestring · can be null

The page title for search engines (domain concept first), at most 70 characters, never a cut word. Null when not generated.

seo_meta_descriptionstring · can be null

The meta description, at most 160 characters; ends at a sentence boundary, or else at a clause or word boundary followed by an ellipsis, never a cut word. Null when not generated.

seo_keywordsarray of string

8-12 domain keywords for meta tags and internal linking.

seo_problem_and_solutionstring · can be null

2-3 paragraphs: the industry problem, the decentralised solution, the Bittensor connection. Null when not generated.

seo_use_casesstring · can be null

3-5 concrete use cases as Markdown bullets (the page's JSON-LD feature list). Null when not generated.

seo_faqsarray of objects

4-6 questions and answers for JSON-LD FAQPage data.

2 fields in seo_faqs (each item)
questionstring

A natural search query phrased as a question.

answerstring

A factual 2-3 sentence answer grounded in the subnet's documentation.

scansarray of objects · can be null

Only with include=scans (absent otherwise): the repository's analysis runs written for this life (the rule analysis uses), newest first, at most 50, without content ("last analysed <date>"). [] without repo.

6 fields in scans (each item)
analyzed_atstring (date-time)

When the analysis ran.

Format: date and time, ISO 8601 in UTC
statusstring

How the analysis run ended. Known values: complete, partial, failed. New values may be added; show an unknown value as is (or generically).

analysis_modelstring · can be null

The model that wrote the analysis. Null when not recorded.

generationinteger · unit id · can be null

Opaque life id (the life's registration identifier), NOT a block height; use life_start_height for the first block. The subnet life the run was stamped with. Null for a run written before stamping, listed because the repository itself is stamped with this life (or, for an ended life, it ran inside the life's time window). Attributed by time when the life's window is known (then this life).

life_start_heightinteger · unit block · can be null

The first block of the life generation names (GET /subnets/generations' start_height): use it, never generation, for heights, ages and dates. Null exactly when generation is null (root without a life, no subnet), or transiently when this API cannot read the lives. Root's generation 0 starts at 0.

is_current_lifeboolean · can be null

Whether generation is the slot's current life (true) or an ended one, re-registered since (false). Null exactly when generation is null, or transiently when this API cannot read the lives. Root's generation 0 is current.

Response headers

10 headers: ETag, Cache-Control, CDN-Cache-Control, X-BTAI-Stale, X-BTAI-Stale-As-Of, RateLimit-Policy, RateLimit, X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset
ETag

Strong validator of this answer. Send it back as If-None-Match: an unchanged answer is a 304 Not Modified without a body. A settled answer's validator leaves the chain head out, so it stays valid as the head moves.

Cache-Control

One of three classes: settled public, max-age=3600, stale-while-revalidate=86400 (pinned to a block that cannot move), slow public, max-age=300, stale-while-revalidate=600 (state that moves in minutes or days) or tip public, max-age=6, stale-while-revalidate=12 (follows the chain head); the label routes are cached until the next nightly copy (public, max-age=<seconds>, stale-while-revalidate=60). Never immutable. no-store on an earlier answer served because the data source failed (X-BTAI-Stale).

CDN-Cache-Control

The same class for shared caches (RFC 9213; browsers ignore it): settled 24 h, slow 300 s, tip 6 s, each with stale-while-revalidate and stale-if-error; no-store with X-BTAI-Stale.

X-BTAI-Stale

Present only when the data source failed to answer (it timed out or was unavailable) and an earlier answer to the same request is served instead of a 503 (stale if error, RFC 5861). Then if-error, with X-BTAI-Stale-As-Of, Age (its seconds since then) and Warning: 110 - "Response is Stale", and Cache-Control: no-store: the answer is never kept as fresh, and a later request reads again. Prices and quotes are never served this way.

X-BTAI-Stale-As-Of

With X-BTAI-Stale: when the earlier answer served was produced, an RFC 3339 UTC time (2026-10-21T12:00:00Z).

RateLimit-Policy

The tier's request quota of the current window (IETF draft), e.g. "tier";q=600.

RateLimit

Requests left in the current window and the seconds until it resets (IETF draft), e.g. "tier";r=597;t=42.

X-RateLimit-Limit

The tier's request quota of the current window.

X-RateLimit-Remaining

Requests left in the current window.

X-RateLimit-Reset

When the current window resets, Unix seconds.

Status codes

200 · 304 · 400 · 401 · 402 · 403 · 404 · 429 · 500 · 502 · 503

304 Not Modified: the If-None-Match validator still matches; no body.

Error codes

The codes this endpoint can return. Each links to what it means and what to do.

From the API's published specification, version 1.165.0.