docs(honcho): add provider switch closeout plan #373

Merged

pdurlej merged 1 commit from codex/honcho-closeout-plan into main

2026-05-18 18:25:32 +02:00

codex commented

2026-05-18 17:54:32 +02:00

Collaborator

Canary status: missing — docs/state/runbook-only plan update; no canary-required paths touched.

Canary Context Pack

Product story

Honcho's Gemma/Ollama switch is live, but the remaining closeout steps were spread across the runbook, issues, runtime evidence, and Oracle guidance. This PR gives the next agent and operator one precise closeout path.

What changed

Added state/cutover/honcho-closeout-plan.md with flights H1-H5: runtime secret bridge, #359 closure, #371 log privacy, #357 embedding-space design, final acceptance.
Updated the Honcho Gemma/Ollama runbook to render LLM_OPENAI_API_KEY from Infisical path /home-platform/apps/honcho:LLM_OPENAI_API_KEY before removing the direct runtime env value.
Updated the prep artifact and STATUS_NOW.md with the corrected Honcho closeout boundary.

Why it changed

Runtime evidence showed OLLAMA_CLOUD_API_KEY and LLM_OPENAI_API_KEY are both available through Infisical, but under different paths. The plan needed to distinguish moving the key source to Infisical from removing OpenAI embedding capability.

Files touched

state/cutover/honcho-closeout-plan.md
runbooks/honcho-ollama-gemma-switch.md
state/cutover/honcho-gemma-ollama-prep.md
state/STATUS_NOW.md

Relevant context

#357 embedding-space migration
#359 Ollama runtime secret access
#371 Honcho raw memory/tool-result logs
Oracle consult honcho-357-embedding-plan, 2026-05-18

Runtime evidence

Metadata-only checks confirmed LLM_OPENAI_API_KEY is present via Infisical at /home-platform/apps/honcho, OLLAMA_CLOUD_API_KEY is present at /home-platform/providers, Honcho containers are healthy, and current soak remains green. No secret values were printed or stored.

Known constraints

OpenAI embedding capability must remain available until #357 has versioned routing, shadow validation, and rollback. The direct runtime env value can be removed only after the Infisical-rendered path is live and embedding smoke passes.

Explicit out-of-scope

No production mutation, no env-file edit on RS2000, no #371 implementation, no #357 schema/code migration, no issue closure.

Requested decision

Approve the closeout plan so the next PRs can execute H1-H4 without re-litigating architecture.

Merge blockers

Any claim that this removes OpenAI embedding capability now, mixes vector spaces, or authorizes production mutation directly.

Spec sources read

runbooks/honcho-ollama-gemma-switch.md — existing Honcho deploy/runbook contract.
state/cutover/honcho-gemma-ollama-prep.md — original LLM switch prep boundary.
state/STATUS_NOW.md — operator-facing current state.
Runtime metadata checks on RS2000 — key presence by Infisical path, no secret values.

Closes #359 only after sanitized evidence is posted separately; this PR does not close it.

Canary status: missing — docs/state/runbook-only plan update; no canary-required paths touched. ## Canary Context Pack ### Product story Honcho's Gemma/Ollama switch is live, but the remaining closeout steps were spread across the runbook, issues, runtime evidence, and Oracle guidance. This PR gives the next agent and operator one precise closeout path. ### What changed - Added `state/cutover/honcho-closeout-plan.md` with flights H1-H5: runtime secret bridge, #359 closure, #371 log privacy, #357 embedding-space design, final acceptance. - Updated the Honcho Gemma/Ollama runbook to render `LLM_OPENAI_API_KEY` from Infisical path `/home-platform/apps/honcho:LLM_OPENAI_API_KEY` before removing the direct runtime env value. - Updated the prep artifact and `STATUS_NOW.md` with the corrected Honcho closeout boundary. ### Why it changed Runtime evidence showed `OLLAMA_CLOUD_API_KEY` and `LLM_OPENAI_API_KEY` are both available through Infisical, but under different paths. The plan needed to distinguish moving the key source to Infisical from removing OpenAI embedding capability. ### Files touched - `state/cutover/honcho-closeout-plan.md` - `runbooks/honcho-ollama-gemma-switch.md` - `state/cutover/honcho-gemma-ollama-prep.md` - `state/STATUS_NOW.md` ### Relevant context - #357 embedding-space migration - #359 Ollama runtime secret access - #371 Honcho raw memory/tool-result logs - Oracle consult `honcho-357-embedding-plan`, 2026-05-18 ### Runtime evidence Metadata-only checks confirmed `LLM_OPENAI_API_KEY` is present via Infisical at `/home-platform/apps/honcho`, `OLLAMA_CLOUD_API_KEY` is present at `/home-platform/providers`, Honcho containers are healthy, and current soak remains green. No secret values were printed or stored. ### Known constraints OpenAI embedding capability must remain available until #357 has versioned routing, shadow validation, and rollback. The direct runtime env value can be removed only after the Infisical-rendered path is live and embedding smoke passes. ### Explicit out-of-scope No production mutation, no env-file edit on RS2000, no #371 implementation, no #357 schema/code migration, no issue closure. ### Requested decision Approve the closeout plan so the next PRs can execute H1-H4 without re-litigating architecture. ### Merge blockers Any claim that this removes OpenAI embedding capability now, mixes vector spaces, or authorizes production mutation directly. ## Spec sources read - `runbooks/honcho-ollama-gemma-switch.md` — existing Honcho deploy/runbook contract. - `state/cutover/honcho-gemma-ollama-prep.md` — original LLM switch prep boundary. - `state/STATUS_NOW.md` — operator-facing current state. - Runtime metadata checks on RS2000 — key presence by Infisical path, no secret values. Closes #359 only after sanitized evidence is posted separately; this PR does not close it.

codex added 1 commit

2026-05-18 17:54:32 +02:00

docs(honcho): add provider switch closeout plan

base-is-main / guard (pull_request) Successful in 1s

Details

canary-required / collect-diff (pull_request) Successful in 4s

Details

patchwarden-pr-sanity / collect-diff (pull_request) Successful in 4s

Details

canary-required / canary (pull_request) Has been skipped

Details

patchwarden-pr-sanity / sanity (pull_request) Successful in 23s