Diff: substrate/tui-corpus-producer
From af1a599 to af1a599
+0 / −0 lines
| Before | After |
|---|---|
| --- | --- |
| schema: foundry-doc-v1 | schema: foundry-doc-v1 |
| title: "TUI as corpus producer" | title: "TUI as corpus producer" |
| slug: tui-corpus-producer | slug: tui-corpus-producer |
| category: substrate | category: substrate |
| type: topic | type: topic |
| quality: complete | quality: complete |
| short_description: "Every terminal interaction with service-slm through the operator TUI is a curated training corpus contribution for the per-tenant adapter." | short_description: "Every terminal interaction with service-slm through the operator TUI is a curated training corpus contribution for the per-tenant adapter." |
| status: active | status: active |
| bcsc_class: public-disclosure-safe | bcsc_class: public-disclosure-safe |
| last_edited: 2026-05-15 | last_edited: 2026-05-15 |
| editor: pointsav-engineering | editor: pointsav-engineering |
| cites: [] | cites: [] |
| references: | references: |
| - id: 1 | - id: 1 |
| text: "Rafailov, R. et al. 'Direct Preference Optimization: Your Language Model is Secretly a Reward Model.' NeurIPS, 2023." | text: "Rafailov, R. et al. 'Direct Preference Optimization: Your Language Model is Secretly a Reward Model.' NeurIPS, 2023." |
| url: "https://arxiv.org/abs/2305.18290" | url: "https://arxiv.org/abs/2305.18290" |
| - id: 2 | - id: 2 |
| text: "Zhou, C. et al. 'LIMA: Less Is More for Alignment.' NeurIPS, 2023." | text: "Zhou, C. et al. 'LIMA: Less Is More for Alignment.' NeurIPS, 2023." |
| url: "https://arxiv.org/abs/2305.11206" | url: "https://arxiv.org/abs/2305.11206" |
| paired_with: tui-corpus-producer.es.md | paired_with: tui-corpus-producer.es.md |
| --- | --- |
| The **TUI-as-Corpus-Producer** pattern designates the operator terminal interface (`slm-cli`) as a primary source of high-quality training data for the per-tenant model adapter. Every interaction with the Doorman through this interface is a curated corpus contribution. | The **TUI-as-Corpus-Producer** pattern designates the operator terminal interface (`slm-cli`) as a primary source of high-quality training data for the per-tenant model adapter. Every interaction with the Doorman through this interface is a curated corpus contribution. |
| ## Why terminal interactions are high-quality training data | ## Why terminal interactions are high-quality training data |
| Three properties distinguish system administration and IT-support interactions from general training data: | Three properties distinguish system administration and IT-support interactions from general training data: |
| **Verifiable ground truth.** When an operator follows AI advice — running a suggested command, applying a proposed configuration change — the system either recovers or it does not. Other domains such as creative writing or strategic reasoning lack this immediate-feedback property. IT-support has it by default. The operator knows immediately whether the response was correct. | **Verifiable ground truth.** When an operator follows AI advice — running a suggested command, applying a proposed configuration change — the system either recovers or it does not. Other domains such as creative writing or strategic reasoning lack this immediate-feedback property. IT-support has it by default. The operator knows immediately whether the response was correct. |
| **Narrow domain.** Archive operations, system conventions, and customer-specific workflow vocabulary form a bounded command set and failure-mode space. Models train more efficiently on bounded domains than on general corpora because the signal-to-noise ratio is higher. | **Narrow domain.** Archive operations, system conventions, and customer-specific workflow vocabulary form a bounded command set and failure-mode space. Models train more efficiently on bounded domains than on general corpora because the signal-to-noise ratio is higher. |
| **Domain-expert feedback.** The operator issuing a verdict is the person who knows whether the response was correct — not a proxy labeler separated from the actual work. Published reinforcement-learning-from-human-feedback literature consistently reports that high-quality verdict-signed interaction tuples train an order of magnitude more efficiently than observation-only tuples. [^1] | **Domain-expert feedback.** The operator issuing a verdict is the person who knows whether the response was correct — not a proxy labeler separated from the actual work. Published reinforcement-learning-from-human-feedback literature consistently reports that high-quality verdict-signed interaction tuples train an order of magnitude more efficiently than observation-only tuples. [^1] |
| ## The /feedback mechanism | ## The /feedback mechanism |
| After every assistant response in the TUI, the operator is offered three explicit verdicts: | After every assistant response in the TUI, the operator is offered three explicit verdicts: |
| **Good.** The response was correct and useful. The tuple is flagged as a positive direct preference optimisation example. | **Good.** The response was correct and useful. The tuple is flagged as a positive direct preference optimisation example. |
| **Refine.** The response was close but needed adjustment. The operator provides a correction inline; the tuple captures the response-and-refinement pair as training signal. | **Refine.** The response was close but needed adjustment. The operator provides a correction inline; the tuple captures the response-and-refinement pair as training signal. |
| **Bad.** The response was wrong. The tuple is flagged as a negative direct preference optimisation example. | **Bad.** The response was wrong. The tuple is flagged as a negative direct preference optimisation example. |
| If the operator dismisses without providing a verdict, the tuple is captured as unsigned and contributes to supervised fine-tuning but not to direct preference optimisation. | If the operator dismisses without providing a verdict, the tuple is captured as unsigned and contributes to supervised fine-tuning but not to direct preference optimisation. |
| ## Adapter quality budget | ## Adapter quality budget |
| Published fine-tuning literature suggests 200 to 500 high-quality verdict-signed interactions are sufficient for a first adapter training cycle in a narrow domain. [^2] The platform's intended sequence for each tenant is: accumulate signed interactions from dogfood operations, train the first per-tenant adapter, apply a validation quality gate, and promote the adapter to the deployment. Each subsequent training cycle incorporates additional interactions, progressively tuning the adapter to the customer's specific environment — their systemd units, their seed taxonomy, their workflow vocabulary. | Published fine-tuning literature suggests 200 to 500 high-quality verdict-signed interactions are sufficient for a first adapter training cycle in a narrow domain. [^2] The platform's intended sequence for each tenant is: accumulate signed interactions from dogfood operations, train the first per-tenant adapter, apply a validation quality gate, and promote the adapter to the deployment. Each subsequent training cycle incorporates additional interactions, progressively tuning the adapter to the customer's specific environment — their systemd units, their seed taxonomy, their workflow vocabulary. |
| ## Per-tenant adapter ownership | ## Per-tenant adapter ownership |
| The corpus produced by a customer's operators trains that customer's adapter, not a general adapter. Per the [[customer-owned-graph-ip]] convention, the trained adapter weights are the customer's property. The platform distributes the model architecture and the training pipeline; the customer retains the trained adapter that results. | The corpus produced by a customer's operators trains that customer's adapter, not a general adapter. Per the [[customer-owned-graph-ip]] convention, the trained adapter weights are the customer's property. The platform distributes the model architecture and the training pipeline; the customer retains the trained adapter that results. |
| ## Verdict capture discipline | ## Verdict capture discipline |
| Some terminal sessions should not contribute to the training corpus: test sessions initiated with a no-corpus flag, sessions interrupted by unavailable tiers before completion, and sessions using forced-tier debug mode are audit-logged but excluded from normal training data. The boundary between operational corpus and test corpus is enforced at the Doorman's verdict intake endpoint. | Some terminal sessions should not contribute to the training corpus: test sessions initiated with a no-corpus flag, sessions interrupted by unavailable tiers before completion, and sessions using forced-tier debug mode are audit-logged but excluded from normal training data. The boundary between operational corpus and test corpus is enforced at the Doorman's verdict intake endpoint. |
| ## See also | ## See also |
| - [[single-boundary-compute-discipline]] — the TUI never calls inference tiers directly; all calls route through the Doorman | - [[single-boundary-compute-discipline]] — the TUI never calls inference tiers directly; all calls route through the Doorman |
| - [[customer-owned-graph-ip]] — per-tenant adapter weights are the customer's intellectual property | - [[customer-owned-graph-ip]] — per-tenant adapter weights are the customer's intellectual property |
| - [[knowledge-graph-grounded-apprenticeship]] — training tuples carry graph context when the Doorman grounds the request | - [[knowledge-graph-grounded-apprenticeship]] — training tuples carry graph context when the Doorman grounds the request |