product- Lugemi Baobab
speech_clarity- spectral_gate_v2
national_models
- baobab-twi
- baobab-ga
- baobab-ha
- baobab-ee
- baobab-fante
- baobab-yo
- baobab-fil
- baobab-fr-sn
- baobab-fr-ci
- baobab-ig
- baobab-ha-ng
- baobab-sw
- baobab-am
- baobab-af
- baobab-zu
apis
baobab_twi- GET /v1/models/baobab-twi
twi_national- GET /v1/baobab/twi/national
twi_research- GET /v1/baobab/twi/research
twi_unlock- GET /v1/baobab/twi/unlock
twi_slo- GET /v1/baobab/twi/slo
baobab_ga- GET /v1/models/baobab-ga
ga_national- GET /v1/baobab/ga/national
ga_research- GET /v1/baobab/ga/research
ga_unlock- GET /v1/baobab/ga/unlock
ga_slo- GET /v1/baobab/ga/slo
baobab_ha- GET /v1/models/baobab-ha
ha_national- GET /v1/baobab/ha/national
ha_research- GET /v1/baobab/ha/research
ha_unlock- GET /v1/baobab/ha/unlock
ha_slo- GET /v1/baobab/ha/slo
baobab_ee- GET /v1/models/baobab-ee
ee_national- GET /v1/baobab/ee/national
ee_research- GET /v1/baobab/ee/research
ee_unlock- GET /v1/baobab/ee/unlock
ee_slo- GET /v1/baobab/ee/slo
baobab_fante- GET /v1/models/baobab-fante
fante_national- GET /v1/baobab/fante/national
fante_research- GET /v1/baobab/fante/research
fante_unlock- GET /v1/baobab/fante/unlock
fante_slo- GET /v1/baobab/fante/slo
baobab_yo- GET /v1/models/baobab-yo
yo_national- GET /v1/baobab/yo/national
yo_research- GET /v1/baobab/yo/research
yo_unlock- GET /v1/baobab/yo/unlock
yo_slo- GET /v1/baobab/yo/slo
baobab_fil- GET /v1/models/baobab-fil
fil_national- GET /v1/baobab/fil/national
fil_research- GET /v1/baobab/fil/research
fil_unlock- GET /v1/baobab/fil/unlock
fil_slo- GET /v1/baobab/fil/slo
baobab_fr_sn- GET /v1/models/baobab-fr-sn
fr_sn_national- GET /v1/baobab/fr-sn/national
fr_sn_research- GET /v1/baobab/fr-sn/research
fr_sn_unlock- GET /v1/baobab/fr-sn/unlock
fr_sn_slo- GET /v1/baobab/fr-sn/slo
baobab_fr_ci- GET /v1/models/baobab-fr-ci
fr_ci_national- GET /v1/baobab/fr-ci/national
fr_ci_research- GET /v1/baobab/fr-ci/research
fr_ci_unlock- GET /v1/baobab/fr-ci/unlock
fr_ci_slo- GET /v1/baobab/fr-ci/slo
baobab_ig- GET /v1/models/baobab-ig
ig_national- GET /v1/baobab/ig/national
ig_research- GET /v1/baobab/ig/research
ig_unlock- GET /v1/baobab/ig/unlock
ig_slo- GET /v1/baobab/ig/slo
baobab_ha_ng- GET /v1/models/baobab-ha-ng
ha_ng_national- GET /v1/baobab/ha-ng/national
ha_ng_research- GET /v1/baobab/ha-ng/research
ha_ng_unlock- GET /v1/baobab/ha-ng/unlock
ha_ng_slo- GET /v1/baobab/ha-ng/slo
baobab_sw- GET /v1/models/baobab-sw
sw_national- GET /v1/baobab/sw/national
sw_research- GET /v1/baobab/sw/research
sw_unlock- GET /v1/baobab/sw/unlock
sw_slo- GET /v1/baobab/sw/slo
baobab_am- GET /v1/models/baobab-am
am_national- GET /v1/baobab/am/national
am_research- GET /v1/baobab/am/research
am_unlock- GET /v1/baobab/am/unlock
am_slo- GET /v1/baobab/am/slo
baobab_af- GET /v1/models/baobab-af
af_national- GET /v1/baobab/af/national
af_research- GET /v1/baobab/af/research
af_unlock- GET /v1/baobab/af/unlock
af_slo- GET /v1/baobab/af/slo
baobab_zu- GET /v1/models/baobab-zu
zu_national- GET /v1/baobab/zu/national
zu_research- GET /v1/baobab/zu/research
zu_unlock- GET /v1/baobab/zu/unlock
zu_slo- GET /v1/baobab/zu/slo
catalogue- GET /v1/baobab/models
translate- POST /v1/translate
demo_speech- POST /v1/demo/speech
residency- GET /v1/residency
honesty
twi
phrase_table_only- true
phrase_coverage_expanded_phase_b- true
national_api_tiers_phase_c- true
phase_d_research_shipped- true
post_d_national_readiness- true
deferred_slice_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
asante_akuapem_interchangeable- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
bare_ak_defaults_to- ak-gh-asante
bare_ak_warning- Bare `ak` / `tw` / `twi` resolves to Asante (`ak-gh-asante`). Pass `ak-gh-akuapem` or `ak-akuapem` for Akuapem. Varieties are never interchangeable. Shared Akan checkpoints + orthography — not exclusive finetunes.
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_asante
id- community_consent_asante
required- true
met- false
evidence_required- Signed Asante contributor agreements reconfirming training use
description- Asante community speakers must reconfirm training export consent.
community_consent_akuapem
id- community_consent_akuapem
required- true
met- false
evidence_required- Signed Akuapem contributor agreements reconfirming training use
description- Akuapem community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
dual_variety_native_review
id- dual_variety_native_review
required- true
met- false
evidence_required- Paid Asante + Akuapem native review records with trainingEligible fields
description- Native review must cover both varieties separately before any train export.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- nllb-200-distilled-600M
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
nllb_license- CC-BY-NC
adapter_status- disabled_until_checkpoint
checkpoint_shipped- false
remote_mt_configured- false
remote_mt_env- null
baobab_twi_nmt_url_configured- false
baobab_twi_nmt_env- null
baobab_twi_nmt_live- false
separate_variety_adapters- false
note- Baobab Twi NMT remote not live. Phrase/lexical packs remain the fail-closed default. Optional BAOBAB_TWI_NMT_URL / LUGEMI_MT_URL when set. NLLB base is CC-BY-NC.
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Accra
affinity_country- GH
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Accra (af-west-1) is the Ghana affinity target; live serving for verbalab + lugemi-asante + baobab-twi-mt is jnb today. No Accra GPU island is deployed. Ministry requests record X-Lugemi-Residency-Affinity soft pin.
licenses
mms_tts_aka- CC-BY-NC-4.0
whisper_twi_v2- Apache-2.0
nllb_200_distilled- CC-BY-NC
mms_commercial_redistribution- restricted_nc
nllb_commercial_redistribution- restricted_nc
note- MMS TTS and NLLB distilled are CC-BY-NC — commercial redistribution restricted. Whisper-twi-v2 is Apache-2.0. neural_nmt true only when NMT health probe succeeds.
deferred
- Commercially cleared NMT + variety-exclusive adapters (NLLB NC live path ≠ commercial clear)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative Twi LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes (recipe shipped; shared checkpoints live)
- trainingEligible unlock evidence (checklist shipped; flag locked false)
ga
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
dangme_interchangeable- false
irish_ga_interchangeable- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- gaa
bare_ga_warning- Irish `ga` / `ga-IE` is not Gã. Pass ISO 639-3 `gaa` or `gaa-GH` / variety `gaa-gh`. Dangme (`ada`) is never interchangeable.
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent. Irish ga is never Ga.
gates
community_consent_accra_ga
id- community_consent_accra_ga
required- true
met- false
evidence_required- Signed Accra Ga contributor agreements reconfirming training use
description- Accra / Ga Mashie community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
accra_ga_native_review
id- accra_ga_native_review
required- true
met- false
evidence_required- Paid Accra Ga native review records with trainingEligible fields
description- Native review must cover Accra Ga before any train export.
dangme_not_conflated
id- dangme_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving Dangme (ada) excluded from Ga train exports
description- Dangme is a separate language — never conflated with Gã training data.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_ga_nmt_url_configured- false
baobab_ga_nmt_env- null
separate_variety_adapters- false
note- Ga MT is phrase/lexical only (en↔gaa). No baobab-ga-mt NLLB service. Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔gaa
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Accra
affinity_country- GH
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Accra (af-west-1) is the Ghana affinity target; live serving for verbalab + lugemi-ga is jnb today. No Accra GPU island is deployed. Ministry requests record X-Lugemi-Residency-Affinity soft pin.
licenses
ga_corpus_tts- CC-BY-NC-4.0
whisper_medium_ga- see_model_card
mms_tts_gaa- absent
note- Corpus TTS is CC-BY-NC-4.0. Meta did not publish mms-tts-gaa. neural_nmt stays false — no Ga NLLB remote shipped.
deferred
- Commercially cleared neural Ga NMT (phrase MT is the live fail-closed path)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative Ga LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared Accra checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
ha
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- hau
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_ha
id- community_consent_ha
required- true
met- false
evidence_required- Signed Accra / Kumasi Zongo contributor agreements reconfirming training use
description- Accra / Kumasi Zongo community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
ha_native_review
id- ha_native_review
required- true
met- false
evidence_required- Paid Hausa (Ghanaian Zongo) native review records with trainingEligible fields
description- Native review must cover Hausa (Ghanaian Zongo) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Kano (ha-ng-kano) is a secondary Nigerian variety — not the Ghana national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_ha_nmt_url_configured- false
baobab_ha_nmt_env- null
separate_variety_adapters- false
note- Baobab Hausa MT is phrase/lexical only (en↔ha). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔ha
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Accra
affinity_country- GH
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Accra (af-west-1) is the Ghana affinity target; live serving for verbalab + lugemi-hausa is jnb today. No Accra GPU island is deployed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
ee
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- ewe
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_ee
id- community_consent_ee
required- true
met- false
evidence_required- Signed Anlo / Ho / Volta contributor agreements reconfirming training use
description- Anlo / Ho / Volta community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
ee_native_review
id- ee_native_review
required- true
met- false
evidence_required- Paid Ewe (Anlo / Volta) native review records with trainingEligible fields
description- Native review must cover Ewe (Anlo / Volta) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Tongu and other Ewe varieties are not interchangeable with Anlo national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_ee_nmt_url_configured- false
baobab_ee_nmt_env- null
separate_variety_adapters- false
note- Baobab Ewe MT is phrase/lexical only (en↔ee). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔ee
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Accra
affinity_country- GH
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Accra (af-west-1) is the Ghana affinity target; live serving for verbalab + lugemi-ewe is jnb today. No Accra GPU island is deployed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
fante
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- fat
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_fante
id- community_consent_fante
required- true
met- false
evidence_required- Signed Cape Coast / Central Region contributor agreements reconfirming training use
description- Cape Coast / Central Region community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
fante_native_review
id- fante_native_review
required- true
met- false
evidence_required- Paid Fante / Mfantse (Cape Coast) native review records with trainingEligible fields
description- Native review must cover Fante / Mfantse (Cape Coast) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Asante and Akuapem are separate Akan varieties under baobab-twi — never interchangeable with Fante.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_fante_nmt_url_configured- false
baobab_fante_nmt_env- null
separate_variety_adapters- false
note- Baobab Fante MT is phrase/lexical only (en↔ak (Fante orthography honesty; shared Akan phrase table)). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔ak (Fante orthography honesty; shared Akan phrase table)
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Accra
affinity_country- GH
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Accra (af-west-1) is the Ghana affinity target; live serving for verbalab + lugemi-asante is jnb today. No Accra GPU island is deployed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
yo
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- yor
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_yo
id- community_consent_yo
required- true
met- false
evidence_required- Signed Lagos / Southwest Nigeria contributor agreements reconfirming training use
description- Lagos / Southwest Nigeria community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
yo_native_review
id- yo_native_review
required- true
met- false
evidence_required- Paid Yoruba (Lagos / Oyo) native review records with trainingEligible fields
description- Native review must cover Yoruba (Lagos / Oyo) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Egba (yo-ng-egba) is a secondary Nigerian variety — not the Ghana national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_yo_nmt_url_configured- false
baobab_yo_nmt_env- null
separate_variety_adapters- false
note- Baobab Yoruba MT is phrase/lexical only (en↔yo). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔yo
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Lagos
affinity_country- NG
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Lagos is the Nigeria residency preference; Africa-west affinity pin remains af-west-1. Live serving for verbalab + lugemi-yoruba is jnb today. No Accra GPU island is deployed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
fil
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- tgl
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_yo
id- community_consent_yo
required- true
met- false
evidence_required- Signed Metro Manila / Philippines contributor agreements reconfirming training use
description- Metro Manila / Philippines community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
yo_native_review
id- yo_native_review
required- true
met- false
evidence_required- Paid Tagalog (Manila / Filipino) native review records with trainingEligible fields
description- Native review must cover Tagalog (Manila / Filipino) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Batangas (fil-ph-batangas) is a secondary Philippines variety — not the Manila national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_fil_nmt_url_configured- false
baobab_fil_nmt_env- null
separate_variety_adapters- false
note- Baobab Tagalog MT is phrase/lexical only (en↔fil / en↔tl). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔fil
- en↔tl
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Manila
affinity_country- NG
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Manila is the Philippines residency preference; Africa-west affinity pin remains af-west-1. Live serving for verbalab + lugemi-tagalog is jnb today. No Accra GPU island is deployed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
fr_sn
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- fr
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_yo
id- community_consent_yo
required- true
met- false
evidence_required- Signed Dakar / Senegal contributor agreements reconfirming training use
description- Dakar / Senegal community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
yo_native_review
id- yo_native_review
required- true
met- false
evidence_required- Paid Senegalese French (Dakar / Senegalese French) native review records with trainingEligible fields
description- Native review must cover Senegalese French (Dakar / Senegalese French) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- France French (fr-FR) and Ivorian French (fr-CI) are not interchangeable with Dakar Senegalese French — not the national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_fr_sn_nmt_url_configured- false
baobab_fr_sn_nmt_env- null
separate_variety_adapters- false
note- Baobab Senegalese French MT is phrase/lexical only (en↔fr). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔fr
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Dakar
affinity_country- SN
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Dakar is the Senegal residency preference; Africa-west affinity pin remains af-west-1. Live serving for verbalab is jnb today; Senegalese French neural speech uses the optional shared Echo worker (no dedicated always-on Fly app). No Accra GPU island is deployed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
fr_ci
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- fr
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_yo
id- community_consent_yo
required- true
met- false
evidence_required- Signed Abidjan / Côte d'Ivoire contributor agreements reconfirming training use
description- Abidjan / Côte d'Ivoire community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
yo_native_review
id- yo_native_review
required- true
met- false
evidence_required- Paid Ivorian French (Abidjan / Ivorian French) native review records with trainingEligible fields
description- Native review must cover Ivorian French (Abidjan / Ivorian French) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- France French (fr-FR) and Senegalese French (fr-SN) are not interchangeable with Abidjan Ivorian French — not the national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_fr_ci_nmt_url_configured- false
baobab_fr_ci_nmt_env- null
separate_variety_adapters- false
note- Baobab Ivorian French MT is phrase/lexical only (en↔fr). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔fr
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Abidjan
affinity_country- CI
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Abidjan is the Côte d'Ivoire residency preference; Africa-west affinity pin remains af-west-1. Live serving for verbalab is jnb today; Ivorian French neural speech uses the optional shared Echo worker (no dedicated always-on Fly app). No Accra GPU island is deployed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
ig
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- ibo
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_yo
id- community_consent_yo
required- true
met- false
evidence_required- Signed Enugu / Southeast Nigeria contributor agreements reconfirming training use
description- Enugu / Southeast Nigeria community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
yo_native_review
id- yo_native_review
required- true
met- false
evidence_required- Paid Igbo (Enugu / Central) native review records with trainingEligible fields
description- Native review must cover Igbo (Enugu / Central) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Owerri (ig-ng-owerri) is a secondary Nigerian variety — not the Nigeria regional national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_ig_nmt_url_configured- false
baobab_ig_nmt_env- null
separate_variety_adapters- false
note- Baobab Igbo MT is phrase/lexical only (en↔ig). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔ig
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Enugu
affinity_country- NG
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Enugu is the Nigeria residency preference; Africa-west affinity pin remains af-west-1. Live serving for verbalab + lugemi-igbo is jnb today. No Accra GPU island is deployed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
ha_ng
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- hau
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_ha_ng
id- community_consent_ha_ng
required- true
met- false
evidence_required- Signed Kano / Northern Nigeria contributor agreements reconfirming training use
description- Kano / Northern Nigeria community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
ha_ng_native_review
id- ha_ng_native_review
required- true
met- false
evidence_required- Paid Hausa (Kano / Nigeria) native review records with trainingEligible fields
description- Native review must cover Hausa (Kano / Nigeria) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Katsina (ha-ng-katsina) is a secondary Nigerian variety — not the Nigeria regional national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_ha_ng_nmt_url_configured- false
baobab_ha_ng_nmt_env- null
separate_variety_adapters- false
note- Baobab Nigeria Hausa (Kano) MT is phrase/lexical only (en↔ha). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔ha
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-west-1
affinity_city- Kano
affinity_country- NG
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-west-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/ACCRA_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Kano is the Nigeria residency preference; Africa-west affinity pin remains af-west-1. Live serving for verbalab + lugemi-hausa is jnb today. No Accra GPU island is deployed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Accra GPU island / hard Ghana geo-fence (affinity af-west-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
sw
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- swh
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_sw
id- community_consent_sw
required- true
met- false
evidence_required- Signed Nairobi / Kenya contributor agreements reconfirming training use
description- Nairobi / Kenya community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
sw_native_review
id- sw_native_review
required- true
met- false
evidence_required- Paid Kiswahili (Nairobi / Kenyan) native review records with trainingEligible fields
description- Native review must cover Kiswahili (Nairobi / Kenyan) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Coastal Tanzania (sw-tz-coastal) is a secondary East African variety — not the Kenya national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_yo_nmt_url_configured- false
baobab_yo_nmt_env- null
separate_variety_adapters- false
note- Baobab Kiswahili MT is phrase/lexical only (en↔sw). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔sw
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-east-1
affinity_city- Nairobi
affinity_country- KE
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-east-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/NAIROBI_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Nairobi is the Kenya residency preference; Africa-east affinity pin remains af-east-1. Live serving for verbalab + lugemi-swahili is jnb today. No Nairobi GPU island fiction is claimed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Nairobi GPU island / hard Kenya geo-fence (affinity af-east-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
am
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- amh
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_sw
id- community_consent_sw
required- true
met- false
evidence_required- Signed Addis Ababa / Ethiopia contributor agreements reconfirming training use
description- Addis Ababa / Ethiopia community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
sw_native_review
id- sw_native_review
required- true
met- false
evidence_required- Paid Amharic (Addis Ababa / Ethiopian) native review records with trainingEligible fields
description- Native review must cover Amharic (Addis Ababa / Ethiopian) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Tigrinya (ti-et) is a secondary East African variety — not the Ethiopia national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_yo_nmt_url_configured- false
baobab_yo_nmt_env- null
separate_variety_adapters- false
note- Baobab Amharic MT is phrase/lexical only (en↔am). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔am
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-east-1
affinity_city- Addis Ababa
affinity_country- ET
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-east-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/NAIROBI_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Addis Ababa is the Ethiopia residency preference; Africa-east affinity pin remains af-east-1. Live serving for verbalab + lugemi-amharic is jnb today. No Addis GPU island fiction is claimed.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Addis GPU island / hard Ethiopia geo-fence (affinity af-east-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
af
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- afr
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_sw
id- community_consent_sw
required- true
met- false
evidence_required- Signed Cape Town / South Africa contributor agreements reconfirming training use
description- Cape Town / South Africa community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
sw_native_review
id- sw_native_review
required- true
met- false
evidence_required- Paid Afrikaans (Cape Town / Standard Afrikaans) native review records with trainingEligible fields
description- Native review must cover Afrikaans (Cape Town / Standard Afrikaans) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Kaaps (ti-et) is a secondary East African variety — not the South Africa national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_yo_nmt_url_configured- false
baobab_yo_nmt_env- null
separate_variety_adapters- false
note- Baobab Afrikaans MT is phrase/lexical only (en↔af). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔af
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-south-1
affinity_city- Cape Town
affinity_country- ZA
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-south-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/JNB_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Cape Town is the South Africa residency preference; Africa-south affinity pin is af-south-1. Live serving for verbalab + lugemi-afrikaans is jnb today. No fictional Cape Town GPU island is claimed live beyond jnb serving.
licenses
tts- CC-BY-4.0
asr- see_model_card
note- TTS is CC-BY-4.0 (Simba). neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Cape Town / JNB GPU island / hard South Africa geo-fence (affinity af-south-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
zu
phrase_table_only- true
national_api_tiers- true
research_slot_shipped- true
neural_speech_live- false
neural_nmt- false
generative_inference_live- false
foundation_llm_weights- false
ships_trained_baobab_foundation- false
variety_exclusive_speech_finetunes- false
trainingEligible- false
iso_code- zul
audio_clean- spectral_gate_v2
training_eligible_unlock
trainingEligible- false
status- locked
unlock_doc- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
note- trainingEligible remains false until all gates are met with durable evidence. Do not invent consent.
gates
community_consent_sw
id- community_consent_sw
required- true
met- false
evidence_required- Signed Durban / South Africa contributor agreements reconfirming training use
description- Durban / South Africa community speakers must reconfirm training export consent.
held_out_speaker_lists
id- held_out_speaker_lists
required- true
met- false
evidence_required- Published held-out speaker IDs never present in train splits
description- Held-out evaluation speakers must never appear in training exports.
sw_native_review
id- sw_native_review
required- true
met- false
evidence_required- Paid isiZulu (Durban / KwaZulu-Natal isiZulu) native review records with trainingEligible fields
description- Native review must cover isiZulu (Durban / KwaZulu-Natal isiZulu) before any train export.
variety_not_conflated
id- variety_not_conflated
required- true
met- false
evidence_required- Corpus metadata proving non-interchangeable varieties excluded from train exports
description- Johannesburg (ti-et) is a secondary East African variety — not the South Africa national default.
rights_clear_for_export
id- rights_clear_for_export
required- true
met- false
evidence_required- Legal memo clearing commercial vs research split for export corpus
description- IP / commercial rights must clear before trainingEligible flips.
nmt
neural_nmt- false
disclosed_base_named- null
disclosed_base_weights_in_repo- false
commercial_license_cleared- false
adapter_status- phrase_table_only
checkpoint_shipped- false
phrase_table_live- true
remote_mt_configured- false
remote_mt_env- null
baobab_yo_nmt_url_configured- false
baobab_yo_nmt_env- null
separate_variety_adapters- false
note- Baobab isiZulu MT is phrase/lexical only (en↔zu). Fail-closed when uncovered. neural_nmt false.
phrase_table_pairs
- en↔zu
residency
serving_today- jnb
serving_today_label- Africa-primary Fly region Johannesburg (jnb)
affinity_target- af-south-1
affinity_city- Durban
affinity_country- ZA
accra_gpu_island- false
accra_gpu_island_status- deferred
hard_geo_fence- false
preference_model- affinity_not_hard_fence
ministry_affinity_pin- af-south-1
affinity_enforcement- soft_preference_recorded
accra_island_runbook- docs/baobab/JNB_RESIDENCY.md
note- Inference prefers matching person/org residency to nearby model hosting when ready — affinity, not a hard geo-fence. Durban is the South Africa residency preference; Africa-south affinity pin is af-south-1. Live serving for verbalab + lugemi-zulu is jnb today. No fictional Durban GPU island is claimed live beyond jnb serving.
licenses
tts- CC-BY-NC-4.0
asr- see_model_card
note- TTS is CC-BY-NC-4.0. neural_nmt stays false — no dedicated NLLB remote shipped. Clarity path: spectral_gate_v2.
deferred
- Commercially cleared neural NMT (phrase MT is the live fail-closed path)
- Durban GPU island / hard South Africa geo-fence (affinity af-south-1; serving jnb)
- Contractual ministry SLA signature (operational targets shipped)
- Generative LM weights + inference (recipe + gates only)
- Variety-exclusive ASR/TTS finetunes beyond shared checkpoints
- trainingEligible unlock evidence (checklist shipped; flag locked false)
docs
twi
national_model- docs/baobab/TWI_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-twi.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
asante_corridor- en-ak-asante
akuapem_corridor- en-ak-akuapem
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
ga
national_model- docs/baobab/GA_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-ga.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
corridor- en-gaa-accra
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
ha
national_model- docs/baobab/HAUSA_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-ha.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
corridor- en-ha-zongo
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
ee
national_model- docs/baobab/EWE_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-ee.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
corridor- en-ee-anlo
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
fante
national_model- docs/baobab/FANTE_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-fante.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
corridor- en-ak-fante
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
yo
national_model- docs/baobab/YORUBA_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-yo.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
corridor- en-yo-lagos
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
fil
national_model- docs/baobab/TAGALOG_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-fil.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
corridor- en-fil-manila
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
fr_sn
national_model- docs/baobab/FR_SN_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-fr-sn.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
corridor- en-fr-sn
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
fr_ci
national_model- docs/baobab/FR_CI_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-fr-ci.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
corridor- en-fr-ci
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
ig
national_model- docs/baobab/IGBO_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-ig.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
corridor- en-ig-enugu
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
ha_ng
national_model- docs/baobab/HAUSA_NG_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-ha-ng.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
accra_residency- docs/baobab/ACCRA_RESIDENCY.md
corridor- en-ha-kano
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
sw
national_model- docs/baobab/SWAHILI_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-sw.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
nairobi_residency- docs/residency.md
corridor- en-sw-nairobi
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
am
national_model- docs/baobab/AMHARIC_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-am.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
nairobi_residency- docs/residency.md
corridor- en-am-addis
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
af
national_model- docs/baobab/AFRIKAANS_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-af.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
jnb_residency- docs/residency.md
corridor- en-af-cape
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md
zu
national_model- docs/baobab/ZULU_NATIONAL_MODEL.md
model_card- docs/baobab/baobab-zu.MODEL_CARD.md
partnership_open_eval- docs/baobab/PARTNERSHIP_OPEN_EVAL.md
ministry_sla- docs/baobab/MINISTRY_SLA.md
training_eligible_unlock- docs/baobab/TRAINING_ELIGIBLE_UNLOCK.md
speech_finetune_recipe- docs/baobab/SPEECH_FINETUNE_RECIPE.md
generative_lm_recipe- docs/baobab/GENERATIVE_LM_RECIPE.md
jnb_residency- docs/residency.md
corridor- en-zu-durban
language_kit- docs/models/05_LANGUAGE_KIT.md
residency- docs/residency.md