Clarify AI Only

Produces the same preferences.json output but with design_constraints limited to region and compliance, startup_constraints populated, and ai_constraints fully populated. Questions are presented in two progressive batches with an intermediate save — partial answers persist across sessions.

Step 0: Prior Run Check

Check $MIGRATION_DIR/ for existing state:

Case 1 — Completed preferences exist (preferences.json present):

“I found existing migration preferences from a previous run. Would you like to:”

1. Re-use these preferences and skip questions
2. Start fresh and re-answer all questions

• If 1: skip to Step 3 (Validation), proceed with existing file.
• If 2: delete preferences.json, continue to Step 1.

Case 2 — Draft preferences exist (preferences-draft.json present, no preferences.json):

“I found a partial set of answers from a previous session (1 of 2 batches completed). Would you like to:”

1. Resume from where you left off — I’ll pick up the remaining questions
2. Start fresh and re-answer all questions

• If 1: load the draft, skip Batch 1 in Step 2, present Batch 2 directly.
• If 2: delete preferences-draft.json, continue to Step 1.

Case 3 — No prior state: Continue to Step 1.

Step 1: Present AI Detection Summary

AI-Only Migration Detected Your project has AI workloads but no infrastructure artifacts (Terraform, billing). I’ll focus on migrating your AI/LLM calls to AWS Bedrock while your infrastructure stays on GCP.

AI source: [from summary.ai_source] Models detected: [from models[].model_id] Capabilities in use: [from integration.capabilities_summary where true] Integration pattern: [from integration.pattern] via [from integration.primary_sdk] Gateway/router: [from integration.gateway_type, or “None (direct SDK)”]

Step 1.5: Fast-Path Check

If migration-preview.json exists and ai_complexity_signal == "likely_simple" (single model, non-agentic, no multi-provider, no multi-model routing):

“Your AI migration looks straightforward — one model swapping to Bedrock. I only need 5 quick answers to complete your migration plan.”

Present only Q1.5, Q2, Q3, Q4, Q11 (Q1 framework is extracted; Q5 model is extracted; Q6 capabilities are extracted; Q7–Q10 use defaults). Q1.5 (compliance) and Q11 (Activate status) are never dropped from the fast path — “never dropped” means always PRESENTED: the fast-path question set must include them; they are never silently omitted or auto-answered. An explicit user “use defaults for the rest” still applies their documented defaults (compliance → ["unknown"] + report caveat — never a silent “none”, matching full-flow Q2 semantics; Activate → unknown + neutral copy) — that is the sanctioned default path, same as full-flow Q27. After answering, skip directly to Step 3.

If ai_complexity_signal is "standard" or "complex", or migration-preview.json is absent, continue to Step 1.75 (mini assumption sheet), then Step 2.

Step 1.75: Mini Assumption Sheet (before Batch 1)

Before asking anything, present what discovery already answered and what will be assumed, as a compact confirm-or-edit sheet (same contract as the full Clarify wizard’s Step 2.5 — sheet first, questions after the user responds):

Questions resolved on this sheet are not re-asked in the batches below; record them with chosen_by: "extracted" (detected rows) or "default" (assumed rows, sheet-confirmed), and skip their entries in Batch 1/2. Rows the user corrects become chosen_by: "user". Rows with no detection signal stay in the batches as normal questions.

Step 2: Ask Questions in Progressive Batches (Q1–Q11, incl. Q1.5)

Questions are presented in two batches with a save after the first. The user can skip individual questions (defaults applied), say “use defaults for the rest” to apply defaults for all remaining questions and proceed immediately, or answer normally.

Batch 1 — AI Strategy & Setup (Q1–Q5 + Q1.5)

Present with this intro:

Q1 — AI framework or orchestration layer (select all that apply)

Same decision logic, auto-detect signals, and interpretation as Q14 in clarify-ai.md.

Auto-detect: No framework → 1, LiteLLM/OpenRouter/Kong/Apigee → 2, LangChain/LangGraph → 3, CrewAI/AutoGen → 4, OpenAI Agents SDK → 5, MCP/A2A → 6, Vapi/Bland.ai/Retell → 7.

Skip when: integration.pattern, integration.gateway_type, and integration.frameworks together give a definitive answer — including a definitive no-framework signal (pattern: "direct_api" with empty frameworks and null gateway_type → 1). Use extracted values with chosen_by: "extracted" and do not present this question. Ask only when the signals are missing or contradict each other.

1) No framework — direct API calls | 2) LLM router/gateway | 3) LangChain / LangGraph | 4) Multi-agent framework | 5) OpenAI Agents SDK | 6) MCP/A2A | 7) Voice platform

Interpret → ai_framework array. Default: auto-detect, fallback ["direct"].

Q1.5 — Do you have any compliance or regulatory requirements? (select all that apply)

Compliance gates Bedrock regions, models, and logging even though your infrastructure stays on GCP — customer data flows to AWS the moment model calls do. Same answer options and decision logic as Q2 in clarify-global.md; the impacts below are the Bedrock-specific subset that applies on this path.

Even with infrastructure staying on GCP, your prompts and completions will be processed on AWS. Compliance requirements determine which Bedrock regions, models, and configurations are available.

1) None | 2) SOC 2 / ISO 27001 | 3) PCI DSS | 4) HIPAA | 5) FedRAMP / Government | 6) GDPR / Data residency | 7) CCPA / CPRA | 8) I don’t know

(Multiple selections allowed)

Interpret → design_constraints.compliance array (same format as the full flow). An explicit user answer of 1 records ["none"] with chosen_by: "user". Skip/default records ["unknown"] (never a silent “none” — full-flow Q2 semantics: behaves like “none” for service selection) with chosen_by: "default", source: "default:Q1.5" — and append the caveat “Compliance requirements were not confirmed by the user” to metadata.report_caveats[] (create the array if absent) so downstream reports surface it. Cross-check with Q4: a GDPR answer constrains the target region jointly with cross-cloud latency.

Q2 — What matters most for your AI application?

1) Best quality/reasoning | 2) Fastest speed | 3) Lowest cost | 4) Specialized capability (→ Q10) | 5) Balanced | 6) I don’t know

Interpret → ai_priority. Default: 5 → "balanced".

Q3 — Monthly AI spend on OpenAI or Gemini?

1) < $500 | 2) $500–$2K | 3) $2K–$10K | 4) > $10K | 5) Don’t know

Interpret → ai_monthly_spend. Default: 2 → "$500-$2K".

Q4 — Cross-cloud API call concerns

Unique to AI-only: infrastructure stays on GCP while AI calls route to AWS.

1) Latency critical — AI in hot path | 2) Latency acceptable — async/users can wait | 3) Concerned about egress costs | 4) Want to test first — parallel running

Interpret → cross_cloud. Default: 2 → "latency-acceptable".

Q5 — Current model in use?

Establishes baseline Bedrock recommendation. Override hierarchy: Q10 special features > Q2 priority > Q7/Q8 volume/latency > Q5 baseline.

Skip when: models[].model_id is populated in ai-workload-profile.json with confidence ≥ 0.8 (the same threshold as full-flow Q19) — auto-detect with chosen_by: "extracted" and do not present this question. The detected models are already shown in the Step 1 summary. Below 0.8, present the question with the detected model(s) offered as the suggested answer. With 2+ detected models, record ai_model_baseline as an array (one entry per model).

1) Gemini Flash | 2) Gemini Pro | 3) GPT-3.5 Turbo | 4) GPT-4/4 Turbo | 5) GPT-4o | 6) GPT-5.4/Mini/Nano | 7) GPT-5.6 Sol/Terra/Luna | 8) GPT-5/5.x (older) | 9) GPT-5.5/Pro | 10) o-series | 11) Claude (Anthropic SDK) | 12) Other/Multiple | 13) Don’t know

Same-model rows first. GPT-5.6 / 5.5 / 5.4 run on Bedrock, so those sources map to themselves. Cost is ~10% ABOVE OpenAI standard (Bedrock in-region is priced at OpenAI’s data-residency tier), so the case is AWS commitments, governance, and residency — not savings, and not parity. They are bedrock-mantle / Responses-only and in-region only; see references/shared/openai-on-bedrock.md. For sources with no Bedrock equivalent, present both a same-vendor upgrade and a cross-family option rather than pre-picking.

Override examples: GPT-4 + Q2=cost → Haiku; Flash + Q10=extended thinking → Sonnet; GPT-4o + Q10=speech → Nova 2 Sonic; GPT-5.5 + Q2=cost → Sonnet 5.

Interpret → ai_model_baseline. Default: auto-detect, fallback Q2 priority-based.

Batch 1 → Save Draft and Present Batch 2

After the user responds to Batch 1:

1. Interpret all Batch 1 answers (apply interpret rules above; apply defaults for skipped questions).
2. Write $MIGRATION_DIR/preferences-draft.json with Batch 1 answers:

1. Present Batch 2:

“Use defaults for the rest” handling: If the user says this during Batch 1, apply defaults for all unanswered Batch 1 questions and all Batch 2 questions, then skip directly to Step 3. Skip the Batch 1 draft save on this path — assembly happens in the same turn, so a draft would serve no crash-recovery purpose.

Batch 2 — Technical Requirements (Q6–Q11)

Q6 — What input types must the model accept: text only, images (vision), or audio/video?

Skip when: integration.capabilities_summary in ai-workload-profile.json has definitive values for vision AND (speech_to_text or text_to_speech) — derive from capabilities with chosen_by: "extracted" and do not present this question. Only ask if capabilities are unknown or ambiguous (all false with no evidence either way).

1) Text only | 2) Vision required | 3) Audio/Video inputs

Interpret → ai_vision. Default: 1 → no constraint.

Q7 — Monthly AI usage volume

Auto-resolve (skip the question): If openai-usage-profile.json exists with non-zero usage, compute total monthly tokens = Σ usage_by_model[].input_tokens + output_tokens, map to the tiers below (< 1M → "low", 1–10M → "medium", 10–100M → "high", > 100M → "very_high"), record the extraction (chosen_by: "extracted", source: "openai-usage-profile:usage_by_model"), and tell the user: “Resolved from your OpenAI usage data: [N tokens/month → tier].” Ask Q7 only if the profile is absent or partial_window makes the volume unreliable.

1) < 1M tokens | 2) 1–10M | 3) 10–100M | 4) > 100M | 5) Don’t know

Interpret → ai_token_volume: 1 → "low", 2 → "medium", 3 → "high", 4 → "very_high". Default: 2 → "medium".

Q8 — Response speed importance

Present with concrete anchors: Critical = autocomplete/live chat; Important = chat assistant; Flexible = reports/batch.

1) Critical (< 500ms) | 2) Important (< 2s) | 3) Flexible (2–10s)

Interpret → ai_latency. Default: 2 → "important".

Q9 — AI task complexity

Present with concrete examples: Simple = classify/extract/summarize; Moderate = analyze+JSON/few-shot; Complex = multi-turn reasoning/tool use/agentic.

1) Simple | 2) Moderate | 3) Complex

Interpret → ai_complexity. Default: 2 → "moderate".

Q10 — Specialized features needed

Same decision logic as Q17 in clarify-ai.md.

1) Function calling | 2) Ultra-long context (> 300K) | 3) Extended thinking | 4) Prompt caching | 5) RAG optimization | 6) Agentic workflows | 7) Real-time speed | 8) Image generation | 9) Conversational speech | 10) None

Interpret → ai_critical_feature. Default: 10 → no override.

Q11 — Have you applied for AWS Activate credits?

Numbering note: AI-only Q11 ≡ full-flow Q27 (startup programs). It is unrelated to the full flow’s Q11 (Cloud Run spend) or Q11b (Graviton) — the two flows number independently.

Same rationale, eligibility rules, and answer semantics as Q27 in clarify-ai.md. AI-only migrations are exactly the workloads Activate credits offset — Bedrock usage (Claude, Llama, Nova) is credit-eligible. Never infer funding stage or Activate tier from Q3 spend (Q27 rule applies here unchanged).

AWS Activate credits offset Bedrock costs during and after migration — including Claude, Llama, and Nova models. Eligible startups can get $5K–$200K depending on funding stage.

1. Yes — already have AWS Activate credits
2. No — haven’t applied yet (self-funded or pre-VC)
3. No — VC/accelerator-backed but haven’t applied
4. I don’t know

Escalations (adapted to AI-only signals): if ai_monthly_spend is ">$10K", also flag AWS Credits for AI Startups ($200,000+, invite-only — contact your AWS Account Manager). If ai_monthly_spend is "$2K-$10K" or ">$10K" AND the workload is agentic (Q1 includes 4/5/6 or Q10 = 6), also flag AWS Generative AI Accelerator (up to $1M credits, 8-week cohort): aws.amazon.com/startups/generative-ai/accelerator

Interpret → startup_program_status: 1 → "has_credits", 2 → "eligible_founders", 3 → "eligible_portfolio", 4 → "unknown". Default: 4 → "unknown" — downstream artifacts must use neutral Activate copy (both tiers, no “your status: eligible_*”).

Batch 2 Complete

After the user responds to Batch 2, interpret all Batch 2 answers and proceed to Step 3.

Step 3: Assemble and Write preferences.json

Assemble all interpreted answers from both batches into the final file. If preferences-draft.json exists, use it as the base — merge in Batch 2 answers, remove draft-specific metadata fields (draft, batches_completed, batches_remaining), and set metadata.timestamp to the current time.

Write $MIGRATION_DIR/preferences.json:

Schema — AI-only structure:

Each constraint carries the FULL clarify.md field shape — value, chosen_by (user|extracted|default|derived), prompt, design_consequence, and source/question_id per the clarify.md source-field rules. (The short { value, chosen_by } form shown above is an abbreviation, not the schema.) No nulls. All schema rules from clarify.md apply, with two AI-only bindings: metadata.clarify_mode is "fast_path" when Step 1.5 fired, "full" for the two-batch flow; and this flow’s questions_extracted is the full flow’s questions_skipped_extracted (downstream consumers accept both names).

After writing preferences.json, delete $MIGRATION_DIR/preferences-draft.json if it exists.

Step 4: Update Phase Status

Before phase completion, enforce output gate:

• preferences.json must exist.
• preferences.json.metadata.migration_type must equal "ai-only".

If either check fails: STOP and output: “AI-only clarify output validation failed. Fix preferences.json before completing Phase 2.”

Use the Phase Status Update Protocol (read-merge-write) to update .phase-status.json in the same turn as the output message:

• Set phases.clarify to "completed"
• Set current_phase to "design"

Output: “Clarification complete. Proceeding to Phase 3: Design AI Migration Architecture.”