open-research-protocol 0.4.30 → 0.4.32

package/AGENT_INTEGRATION.md

@@ -103,8 +103,10 @@ artifact paths (code/data/proofs/logs/papers).
 - Treat **failed paths** as assets: record dead ends as a `Failed Path Record` with the blocking reason/counterexample and a
   next hook.
 - Resolve disputes by **verification or downgrade**, not argument.
-- Run `orp hygiene --json` before long delegation, after material writeback, before API/remote/paid compute, and when dirty
-  state grows unexpectedly.
+- Run `orp hygiene --json` before long delegation, after material writeback, before remote side effects or unbudgeted paid
+  compute, and when dirty state grows unexpectedly.
+- Do not hard-stop solely because an OpenAI research lane is paid; budgeted ORP research may run when `orp research` spend
+  preflight is within the configured daily cap.
 - Stop long-running expansion while hygiene reports `dirty_unclassified`; classify, refresh generated surfaces, canonicalize
   useful scratch, or write a blocker before continuing.
 - Hygiene is non-destructive: never reset, checkout, or delete files merely to hide dirty state.

package/CHANGELOG.md

@@ -6,6 +6,32 @@ There was no prior in-repo changelog file, so the first formal entry starts
 with the currently shipped `v0.4.4` release and summarizes the full release
 delta reflected in this repo.
 
+## v0.4.32 - 2026-04-25
+
+This release clarifies ORP's paid-work boundary so budgeted OpenAI research is
+not treated as a hard stop solely because it uses paid API calls.
+
+### Changed
+
+- Built-in OpenAI research lanes now require a local spend policy, then use
+  spend preflight as the approval boundary for budgeted provider calls.
+- Generated project context, AGENTS guidance, handoffs, and research docs now
+  distinguish budgeted ORP research from unbudgeted paid compute, purchases,
+  and cap-exceeded provider calls.
+
+## v0.4.31 - 2026-04-25
+
+This release refreshes ORP's OpenAI-backed research lanes and tightens
+workspace tab recency ranking for grouped project sessions.
+
+### Changed
+
+- Updated built-in OpenAI research profiles to use `gpt-5.5` for high-reasoning,
+  web synthesis, and pro-research-style lanes, with Responses API `web_search`
+  and `xhigh` reasoning on deep research passes.
+- Workspace tab reports now rank grouped Codex project tabs by the freshest
+  tracked session update time while keeping same-project sessions together.
+
 ## v0.4.30 - 2026-04-25
 
 This release tightens ORP-managed Codex session tracking so short-lived

package/cli/orp.py

@@ -141,9 +141,16 @@ FRONTIER_TERMINAL_STATUSES = {"complete", "completed", "done", "skipped", "termi
 YOUTUBE_SOURCE_SCHEMA_VERSION = "1.0.0"
 EXCHANGE_REPORT_SCHEMA_VERSION = "1.0.0"
 RESEARCH_RUN_SCHEMA_VERSION = "1.0.0"
+OPENAI_RESEARCH_MODEL = "gpt-5.5"
+OPENAI_DEEP_RESEARCH_MODEL = OPENAI_RESEARCH_MODEL
 SECRET_SPEND_POLICY_SCHEMA_VERSION = "1.0.0"
 RESEARCH_SPEND_LEDGER_SCHEMA_VERSION = "1.0.0"
 PROJECT_CONTEXT_SCHEMA_VERSION = "1.0.0"
+HYGIENE_REMOTE_SPEND_MOMENT = "before remote side effects or unbudgeted paid compute"
+BUDGETED_RESEARCH_SPEND_RULE = (
+    "Do not hard-stop solely because an OpenAI research lane is paid; budgeted ORP research may run when "
+    "`orp research` spend preflight is within the configured daily cap."
+)
 HYGIENE_POLICY_SCHEMA_VERSION = "1.0.0"
 MAINTENANCE_STATE_SCHEMA_VERSION = "1.0.0"
 SCHEDULE_REGISTRY_SCHEMA_VERSION = "1.0.0"
@@ -6919,7 +6926,7 @@ def _default_hygiene_policy() -> dict[str, Any]:
     run_moments = [
         "before long delegation",
         "after material writeback",
-        "before API/remote/paid compute",
+        HYGIENE_REMOTE_SPEND_MOMENT,
         "when dirty state grows unexpectedly",
     ]
     self_healing_policy = [
@@ -10661,23 +10668,23 @@ def _project_research_trigger_policy() -> dict[str, Any]:
                 "moment_id": "thinking_reasoning_high",
                 "calls_api": True,
                 "lane": "openai_reasoning_high",
-                "model": "gpt-5.4",
+                "model": OPENAI_RESEARCH_MODEL,
                 "when": "Use when the directory has a decision gate, route choice, proof strategy, architecture tradeoff, or ambiguous next action.",
             },
             {
                 "moment_id": "web_synthesis",
                 "calls_api": True,
                 "lane": "openai_web_synthesis",
-                "model": "gpt-5.4",
+                "model": OPENAI_RESEARCH_MODEL,
                 "when": "Use when the answer depends on current public facts, external docs, papers, project status, or citations.",
             },
             {
                 "moment_id": "pro_deep_research",
                 "calls_api": True,
                 "lane": "openai_deep_research",
-                "model": "o3-deep-research-2025-06-26",
+                "model": OPENAI_DEEP_RESEARCH_MODEL,
                 "when": "Use only after reasoning/web lanes expose a research-heavy gap, disagreement, source-quality issue, or literature-scale synthesis need.",
-                "capability_note": "Requires an OpenAI organization verified for Deep Research model access.",
+                "capability_note": "Runs GPT-5.5 with background mode, web search, and xhigh reasoning for a pro-research-style pass.",
             },
         ],
         "skip_research_when": [
@@ -10690,6 +10697,11 @@ def _project_research_trigger_policy() -> dict[str, Any]:
             "the project must compare multiple papers, standards, providers, or public claims",
             "the output needs a citation-rich report rather than a short decision memo",
         ],
+        "spend_policy": {
+            "budgeted_provider_calls": "OpenAI research lanes are paid but allowed when executed through ORP with a configured local spend policy and a passing spend preflight.",
+            "hard_stop_boundary": "Stop for missing required spend policy, missing secret, cap-exceeded preflight, unbudgeted provider spend, purchases, or non-ORP paid compute.",
+            "local_enforcement": "keychain spend policy with local_preflight_reservation",
+        },
     }
 
 
@@ -10709,7 +10721,7 @@ def _project_evolution_policy() -> dict[str, Any]:
             "run_moments": [
                 "before long delegation",
                 "after material writeback",
-                "before API/remote/paid compute",
+                HYGIENE_REMOTE_SPEND_MOMENT,
                 "when dirty state grows unexpectedly",
             ],
             "stop_rule": (
@@ -10718,10 +10730,11 @@ def _project_evolution_policy() -> dict[str, Any]:
                 "or write a blocker first."
             ),
             "self_healing_rule": "Non-destructive by default: never reset, checkout, or delete files merely to hide dirty state.",
+            "budgeted_research_spend_rule": BUDGETED_RESEARCH_SPEND_RULE,
         },
         "evolution_loop": [
             "scan authority surfaces",
-            "run worktree hygiene before expansion or remote spend",
+            "run worktree hygiene before expansion, remote side effects, or unbudgeted spend",
             "classify dirty state as canonical, runtime, source/test, docs, scratch, or blocker",
             "classify what is local, public, executable, or human-gated",
             "choose whether reasoning, web synthesis, or deep research is justified",
@@ -10762,9 +10775,10 @@ def _project_context_payload(repo_root: Path, *, source: str) -> dict[str, Any]:
             "run_moments": [
                 "before long delegation",
                 "after material writeback",
-                "before API/remote/paid compute",
+                HYGIENE_REMOTE_SPEND_MOMENT,
                 "when dirty state grows unexpectedly",
             ],
+            "budgeted_research_spend_rule": BUDGETED_RESEARCH_SPEND_RULE,
         },
         "evolution_policy": _project_evolution_policy(),
         "next_actions": [
@@ -10778,6 +10792,7 @@ def _project_context_payload(repo_root: Path, *, source: str) -> dict[str, Any]:
             "This file is ORP process context for the local directory.",
             "It is refreshed as the project evolves and should not be cited as proof or canonical evidence.",
             "Provider research calls remain opt-in through `orp research ask --execute`.",
+            BUDGETED_RESEARCH_SPEND_RULE,
         ],
     }
 
@@ -10910,7 +10925,8 @@ def _init_handoff_template(repo_root: Path, *, default_branch: str, initialized_
         "## Agent Rules\n\n"
         f"- Do not do meaningful implementation work directly on `{default_branch}` unless explicitly allowed.\n"
         "- Create a work branch before substantial edits.\n"
-        "- Run `orp hygiene --json` before long delegation, after material writeback, before API/remote/paid compute, and when dirty state grows unexpectedly.\n"
+        f"- Run `orp hygiene --json` before long delegation, after material writeback, {HYGIENE_REMOTE_SPEND_MOMENT}, and when dirty state grows unexpectedly.\n"
+        f"- {BUDGETED_RESEARCH_SPEND_RULE}\n"
         "- Stop long-running expansion while hygiene reports `dirty_unclassified`; classify, refresh generated surfaces, canonicalize useful scratch, or write a blocker.\n"
         "- Hygiene is non-destructive: never reset, checkout, or delete files merely to hide dirty state.\n"
         "- Create a checkpoint commit after each meaningful completed unit of work.\n"
@@ -11096,7 +11112,8 @@ def _render_agent_guide_block(
             [
                 "- Preserve human notes outside ORP-managed blocks.",
                 "- Use this local file for the project-specific current state, local constraints, and concrete next moves.",
-                "- Run `orp hygiene --json` before long delegation, after material writeback, before API/remote/paid compute, and when dirty state grows unexpectedly.",
+                f"- Run `orp hygiene --json` before long delegation, after material writeback, {HYGIENE_REMOTE_SPEND_MOMENT}, and when dirty state grows unexpectedly.",
+                f"- {BUDGETED_RESEARCH_SPEND_RULE}",
                 "- Stop long-running expansion while hygiene reports `dirty_unclassified`; classify, refresh generated surfaces, canonicalize useful scratch, or write a blocker.",
                 "- Hygiene is non-destructive: never reset, checkout, or delete files merely to hide dirty state.",
             ]
@@ -11514,9 +11531,10 @@ def _agent_policy_payload(
             "run_moments": [
                 "before long delegation",
                 "after material writeback",
-                "before API/remote/paid compute",
+                HYGIENE_REMOTE_SPEND_MOMENT,
                 "when dirty state grows unexpectedly",
             ],
+            "budgeted_research_spend_rule": BUDGETED_RESEARCH_SPEND_RULE,
             "required_self_healing": [
                 "classify dirty paths",
                 "refresh generated surfaces",
@@ -17656,7 +17674,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 "calls_api": True,
                 "secret_alias": "openai-primary",
                 "env_var": "OPENAI_API_KEY",
-                "description": "Call GPT-5.4 with high reasoning to critique and compress the opening research.",
+                "description": f"Call {OPENAI_RESEARCH_MODEL} with high reasoning to critique and compress the opening research.",
             },
             {
                 "moment_id": "think_web_crosscheck",
@@ -17664,7 +17682,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 "calls_api": True,
                 "secret_alias": "openai-primary",
                 "env_var": "OPENAI_API_KEY",
-                "description": "Call GPT-5.4 with high reasoning and web search to verify recency-sensitive claims.",
+                "description": f"Call {OPENAI_RESEARCH_MODEL} with high reasoning and web search to verify recency-sensitive claims.",
             },
             {
                 "moment_id": "think_synthesis",
@@ -17672,7 +17690,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 "calls_api": True,
                 "secret_alias": "openai-primary",
                 "env_var": "OPENAI_API_KEY",
-                "description": "Call GPT-5.4 with high reasoning to resolve disagreements before final research.",
+                "description": f"Call {OPENAI_RESEARCH_MODEL} with high reasoning to resolve disagreements before final research.",
             },
             {
                 "moment_id": "final_deep_research",
@@ -17691,7 +17709,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 "call_moment": "opening_deep_research",
                 "label": "Opening Deep Research",
                 "provider": "openai",
-                "model": "o3-deep-research-2025-06-26",
+                "model": OPENAI_DEEP_RESEARCH_MODEL,
                 "adapter": "openai_responses",
                 "role": (
                     "Initial Deep Research scan. Map the landscape, source families, hard unknowns, "
@@ -17712,9 +17730,11 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 ],
                 "env_var": "OPENAI_API_KEY",
                 "secret_alias": "openai-primary",
+                "spend_policy_required": True,
+                "reasoning_effort": "xhigh",
                 "reasoning_summary": "auto",
                 "web_search": True,
-                "web_search_tool": "web_search_preview",
+                "web_search_tool": "web_search",
                 "background": False,
                 "spend_reserve_usd": 1.5,
                 "max_tool_calls": 40,
@@ -17728,7 +17748,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 "call_moment": "think_after_deep",
                 "label": "Think after Deep Research",
                 "provider": "openai",
-                "model": "gpt-5.4",
+                "model": OPENAI_RESEARCH_MODEL,
                 "adapter": "openai_responses",
                 "role": (
                     "High-reasoning critique of the opening Deep Research output. Compress it into a sharper "
@@ -17748,6 +17768,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 ],
                 "env_var": "OPENAI_API_KEY",
                 "secret_alias": "openai-primary",
+                "spend_policy_required": True,
                 "reasoning_effort": "high",
                 "text_verbosity": "medium",
                 "spend_reserve_usd": 0.5,
@@ -17761,7 +17782,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 "call_moment": "think_web_crosscheck",
                 "label": "Think with web cross-check",
                 "provider": "openai",
-                "model": "gpt-5.4",
+                "model": OPENAI_RESEARCH_MODEL,
                 "adapter": "openai_responses",
                 "role": (
                     "High-reasoning web-search pass. Verify current facts, citations, public claims, "
@@ -17781,6 +17802,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 ],
                 "env_var": "OPENAI_API_KEY",
                 "secret_alias": "openai-primary",
+                "spend_policy_required": True,
                 "reasoning_effort": "high",
                 "text_verbosity": "medium",
                 "web_search": True,
@@ -17799,7 +17821,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 "call_moment": "think_synthesis",
                 "label": "Synthesis thinking pass",
                 "provider": "openai",
-                "model": "gpt-5.4",
+                "model": OPENAI_RESEARCH_MODEL,
                 "adapter": "openai_responses",
                 "role": (
                     "High-reasoning synthesis pass. Reconcile the deep-research map, critique, and web cross-check "
@@ -17818,6 +17840,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 ],
                 "env_var": "OPENAI_API_KEY",
                 "secret_alias": "openai-primary",
+                "spend_policy_required": True,
                 "reasoning_effort": "high",
                 "text_verbosity": "medium",
                 "spend_reserve_usd": 0.5,
@@ -17831,7 +17854,7 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 "call_moment": "final_deep_research",
                 "label": "Final Deep Research",
                 "provider": "openai",
-                "model": "o3-deep-research-2025-06-26",
+                "model": OPENAI_DEEP_RESEARCH_MODEL,
                 "adapter": "openai_responses",
                 "role": (
                     "Final Deep Research pass. Use all prior lane outputs to produce the decisive, source-grounded "
@@ -17851,9 +17874,11 @@ def _research_staged_deep_think_profile(profile_id: str = "deep-think-web-think-
                 ],
                 "env_var": "OPENAI_API_KEY",
                 "secret_alias": "openai-primary",
+                "spend_policy_required": True,
+                "reasoning_effort": "xhigh",
                 "reasoning_summary": "auto",
                 "web_search": True,
-                "web_search_tool": "web_search_preview",
+                "web_search_tool": "web_search",
                 "background": False,
                 "spend_reserve_usd": 1.5,
                 "max_tool_calls": 40,
@@ -17901,7 +17926,7 @@ def _research_default_profile(profile_id: str = "openai-council") -> dict[str, A
                 "calls_api": True,
                 "secret_alias": "openai-primary",
                 "env_var": "OPENAI_API_KEY",
-                "description": "Call GPT-5.4 with high reasoning for the deliberate thinking pass.",
+                "description": f"Call {OPENAI_RESEARCH_MODEL} with high reasoning for the deliberate thinking pass.",
             },
             {
                 "moment_id": "web_synthesis",
@@ -17909,7 +17934,7 @@ def _research_default_profile(profile_id: str = "openai-council") -> dict[str, A
                 "calls_api": True,
                 "secret_alias": "openai-primary",
                 "env_var": "OPENAI_API_KEY",
-                "description": "Call GPT-5.4 with web search for current public evidence and citations.",
+                "description": f"Call {OPENAI_RESEARCH_MODEL} with web search for current public evidence and citations.",
             },
             {
                 "moment_id": "pro_deep_research",
@@ -17926,11 +17951,12 @@ def _research_default_profile(profile_id: str = "openai-council") -> dict[str, A
                 "call_moment": "thinking_reasoning_high",
                 "label": "OpenAI reasoning high",
                 "provider": "openai",
-                "model": "gpt-5.4",
+                "model": OPENAI_RESEARCH_MODEL,
                 "adapter": "openai_responses",
                 "role": "Deliberate high-reasoning pass from the provided context. Think hard, critique assumptions, and produce a decision-oriented answer.",
                 "env_var": "OPENAI_API_KEY",
                 "secret_alias": "openai-primary",
+                "spend_policy_required": True,
                 "reasoning_effort": "high",
                 "text_verbosity": "medium",
                 "spend_reserve_usd": 0.5,
@@ -17941,11 +17967,12 @@ def _research_default_profile(profile_id: str = "openai-council") -> dict[str, A
                 "call_moment": "web_synthesis",
                 "label": "OpenAI web synthesis",
                 "provider": "openai",
-                "model": "gpt-5.4",
+                "model": OPENAI_RESEARCH_MODEL,
                 "adapter": "openai_responses",
                 "role": "Recency-aware synthesis using OpenAI Responses web search with citations.",
                 "env_var": "OPENAI_API_KEY",
                 "secret_alias": "openai-primary",
+                "spend_policy_required": True,
                 "reasoning_effort": "high",
                 "text_verbosity": "medium",
                 "web_search": True,
@@ -17961,14 +17988,16 @@ def _research_default_profile(profile_id: str = "openai-council") -> dict[str, A
                 "call_moment": "pro_deep_research",
                 "label": "OpenAI Pro / Deep Research",
                 "provider": "openai",
-                "model": "o3-deep-research-2025-06-26",
+                "model": OPENAI_DEEP_RESEARCH_MODEL,
                 "adapter": "openai_responses",
                 "role": "Pro Research style long-form investigation. Produce a structured, citation-rich report grounded in public sources.",
                 "env_var": "OPENAI_API_KEY",
                 "secret_alias": "openai-primary",
+                "spend_policy_required": True,
+                "reasoning_effort": "xhigh",
                 "reasoning_summary": "auto",
                 "web_search": True,
-                "web_search_tool": "web_search_preview",
+                "web_search_tool": "web_search",
                 "background": True,
                 "spend_reserve_usd": 3.5,
                 "max_tool_calls": 40,
@@ -18480,6 +18509,7 @@ def _research_openai_spend_preflight(
     provider = str(lane.get("provider", "") or "").strip()
     secret_alias = str(lane.get("secret_alias", "") or "").strip()
     reserve_usd = _research_lane_spend_reserve_usd(lane)
+    spend_policy_required = bool(lane.get("spend_policy_required", False))
     entry, entry_issue = _research_spend_policy_entry_for_lane(lane)
     policy = _normalize_secret_spend_policy(entry.get("spend_policy", {}) if isinstance(entry, dict) else {})
     date_utc = dt.datetime.now(dt.timezone.utc).date().isoformat()
@@ -18493,11 +18523,15 @@ def _research_openai_spend_preflight(
         "ledger_path": str(_research_spend_ledger_path()),
     }
     if not policy:
+        reason = entry_issue or "no spend policy configured for this local keychain entry"
+        if spend_policy_required:
+            reason = f"required spend policy missing: {reason}"
         return {
             **base,
-            "allowed": True,
+            "allowed": not spend_policy_required,
             "policy_source": "",
-            "reason": entry_issue or "no spend policy configured for this local keychain entry",
+            "spend_policy_required": spend_policy_required,
+            "reason": reason,
         }
 
     reserved_today = _research_spend_ledger_today_total(
@@ -18514,6 +18548,7 @@ def _research_openai_spend_preflight(
         **base,
         "allowed": allowed,
         "policy_source": "keychain",
+        "spend_policy_required": spend_policy_required,
         "daily_cap_usd": round(daily_cap_usd, 6),
         "currency": str(policy.get("currency", "USD")).strip() or "USD",
         "reserved_today_usd": reserved_today,
@@ -18914,7 +18949,7 @@ def _research_run_openai_lane(
         }
 
     body: dict[str, Any] = {
-        "model": str(lane.get("model", "gpt-5.4")).strip() or "gpt-5.4",
+        "model": str(lane.get("model", OPENAI_RESEARCH_MODEL)).strip() or OPENAI_RESEARCH_MODEL,
         "input": prompt,
         "background": bool(lane.get("background", False)),
     }

package/docs/AGENT_LOOP.md

@@ -13,10 +13,15 @@ read:
 - Read `llms.txt`.
 - Run `orp about --json`.
 - Run `orp hygiene --json` before long delegation, after material writeback,
-  before API/remote/paid compute, and whenever dirty state grows unexpectedly.
+  before remote side effects or unbudgeted paid compute, and whenever dirty
+  state grows unexpectedly.
   If it reports `dirty_unclassified`, stop long-running expansion and classify
   the paths, refresh generated surfaces, canonicalize useful scratch, or write a
   blocker before continuing.
+- Do not hard-stop solely because an OpenAI research lane is paid. Budgeted ORP
+  research may run when `orp research` spend preflight is within the configured
+  daily cap; stop for missing required spend policy, missing secret, cap
+  exhaustion, purchases, or non-ORP paid compute.
 - If the task benefits from fresh concepting, tasteful interface work, or
   exploratory reframing, run:
   - `orp mode nudge sleek-minimal-progressive --json`

package/docs/ORP_AUTONOMY_PROJECT_COMPILATION_MODEL.md

@@ -91,7 +91,7 @@ A true gate is not "the agent feels uncertain."
 
 A true gate is a boundary like:
 
-- spend or purchase
+- unbudgeted spend or purchase
 - outreach or counterparty contact
 - provider/vendor selection with real consequences
 - legal/oversight/compliance judgment
@@ -196,7 +196,7 @@ Bad candidates for automatic compilation:
 
 - vague strategic narratives with no runnable command
 - tasks that imply counterparty contact
-- tasks that imply money
+- tasks that imply unbudgeted money
 - steps that promote support-only outputs into authority
 
 ## What ORP Should Emit
@@ -230,6 +230,7 @@ The controller benchmark experiment surfaced the exact shape:
 2. compile the remaining pre-outreach tasks
 3. keep drafts unsent
 4. stop only when the next step would actually contact a counterparty or spend
+   outside a configured budget/preflight policy
 5. emit a gate dossier
 6. resume only after the human opens that gate
 

package/docs/RESEARCH_COUNCIL.md

@@ -22,11 +22,11 @@ orp research ask "Where should this system live?" --execute --json
 
 The built-in `openai-council` profile defines three OpenAI API lanes:
 
-- `openai_reasoning_high`: `gpt-5.4` with `reasoning.effort=high` for the deliberate thinking pass.
-- `openai_web_synthesis`: `gpt-5.4` with high reasoning plus Responses API web search for current public evidence and citations.
-- `openai_deep_research`: `o3-deep-research-2025-06-26` with background execution and web search preview for Pro/Deep Research style investigation.
+- `openai_reasoning_high`: `gpt-5.5` with `reasoning.effort=high` for the deliberate thinking pass.
+- `openai_web_synthesis`: `gpt-5.5` with high reasoning plus Responses API web search for current public evidence and citations.
+- `openai_deep_research`: `gpt-5.5` with `reasoning.effort=xhigh`, background execution, and Responses API web search for Pro/Deep Research style investigation.
 
-This follows OpenAI's current model guidance: `gpt-5.4` is the default for general-purpose, coding, reasoning, and agentic workflows; web search is enabled through the Responses API `tools` array when current information is needed; and Deep Research is available through the Responses endpoint with `o3-deep-research-2025-06-26`.
+This follows OpenAI's current model guidance: `gpt-5.5` works best through the Responses API for reasoning and tool workflows; web search is enabled through the Responses API `tools` array when current information is needed; and deeper research-style work should use higher reasoning effort plus background mode.
 
 ## Staged Deep Research Template
 
@@ -102,6 +102,26 @@ printf '%s' '<openai-key>' | orp secrets keychain-add \
   --json
 ```
 
+## Spend Policy
+
+The OpenAI research lanes are paid, but paid does not automatically mean human
+hard stop. ORP treats them as budgeted provider calls when `openai-primary` has
+a local spend policy and the lane passes spend preflight.
+
+Set or update the local daily cap metadata like this:
+
+```bash
+orp secrets keychain-spend-policy openai-primary \
+  --daily-spend-cap-usd 5 \
+  --dashboard-spend-cap-status unconfirmed \
+  --dashboard-url https://platform.openai.com/settings/organization/limits \
+  --json
+```
+
+Live research should stop when the required spend policy is missing, the secret
+is missing, the daily cap would be exceeded, or the work is unbudgeted provider
+spend outside ORP research lanes.
+
 ## Fixtures
 
 Provider outputs can be attached without spending live calls:
@@ -119,7 +139,7 @@ Fixtures are useful when an OpenAI run happened outside ORP, when you are compar
 
 ORP uses the Responses API for these lanes. Useful knobs in profile JSON:
 
-- `model`: for example `gpt-5.4` or `o3-deep-research-2025-06-26`.
+- `model`: for example `gpt-5.5`.
 - `call_moment`: the named research-loop moment when this lane may resolve a key.
 - `reasoning_effort`: `none`, `low`, `medium`, `high`, or `xhigh` for supported models.
 - `reasoning_summary`: `auto` or `detailed` for Deep Research reasoning summaries.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-research-protocol",
-  "version": "0.4.30",
+  "version": "0.4.32",
   "description": "ORP CLI (Open Research Protocol): workspace ledgers, secrets, scheduling, governed execution, and agent-friendly research workflows.",
   "license": "MIT",
   "author": "Fractal Research Group <cody@frg.earth>",

package/packages/orp-workspace-launcher/src/tabs.js

@@ -122,9 +122,9 @@ function buildCodexActivityIndex(tabs = [], options = {}) {
   return activityBySessionId;
 }
 
-function orderTabsByRecentActivity(tabs = [], options = {}) {
+function buildRankedTabs(tabs = [], options = {}) {
   const activityBySessionId = buildCodexActivityIndex(tabs, options);
-  const rankedTabs = tabs.map((tab, originalIndex) => {
+  return tabs.map((tab, originalIndex) => {
     const sessionActivity =
       tab.resumeTool === "codex" && tab.sessionId ? activityBySessionId.get(String(tab.sessionId).toLowerCase()) : null;
     return {
@@ -133,24 +133,40 @@ function orderTabsByRecentActivity(tabs = [], options = {}) {
       activityMs: sessionActivity?.mtimeMs || 0,
     };
   });
+}
+
+function orderTabsByRecentActivity(tabs = [], options = {}) {
+  const rankedTabs = buildRankedTabs(tabs, options);
+  const projects = new Map();
 
-  const projectActivity = new Map();
   for (const ranked of rankedTabs) {
-    const current = projectActivity.get(ranked.tab.path) || 0;
-    projectActivity.set(ranked.tab.path, Math.max(current, ranked.activityMs));
+    const projectPath = ranked.tab.path;
+    if (!projects.has(projectPath)) {
+      projects.set(projectPath, {
+        projectPath,
+        firstIndex: ranked.originalIndex,
+        activityMs: ranked.activityMs,
+        tabs: [],
+      });
+    }
+
+    const project = projects.get(projectPath);
+    project.firstIndex = Math.min(project.firstIndex, ranked.originalIndex);
+    project.activityMs = Math.max(project.activityMs, ranked.activityMs);
+    project.tabs.push(ranked);
   }
 
-  return rankedTabs
-    .sort((left, right) => {
-      const leftProjectActivity = projectActivity.get(left.tab.path) || 0;
-      const rightProjectActivity = projectActivity.get(right.tab.path) || 0;
-      return (
-        rightProjectActivity - leftProjectActivity ||
+  return [...projects.values()]
+    .sort(
+      (left, right) =>
         right.activityMs - left.activityMs ||
-        left.originalIndex - right.originalIndex
-      );
-    })
-    .map((ranked) => ranked.tab);
+        left.firstIndex - right.firstIndex,
+    )
+    .flatMap((project) =>
+      project.tabs
+        .sort((left, right) => right.activityMs - left.activityMs || left.originalIndex - right.originalIndex)
+        .map((ranked) => ranked.tab),
+    );
 }
 
 export function parseWorkspaceTabsArgs(argv = []) {

package/packages/orp-workspace-launcher/test/commands.test.js

@@ -74,12 +74,12 @@ test("buildWorkspaceCommandsReport exposes direct restart commands and exact sav
   assert.equal(report.commandCount, 3);
   assert.equal(report.tabs[0]?.resumeCommand, "codex resume abc-123");
   assert.equal(report.tabs[0]?.restartCommand, "cd '/Volumes/Code_2TB/code/collaboration' && codex resume abc-123");
-  assert.equal(report.tabs[1]?.resumeCommand, "claude resume claude-456");
+  assert.equal(report.tabs[1]?.restartCommand, "cd '/Volumes/Code_2TB/code/collaboration'");
+  assert.equal(report.tabs[2]?.resumeCommand, "claude resume claude-456");
   assert.equal(
-    report.tabs[1]?.restartCommand,
+    report.tabs[2]?.restartCommand,
     "cd '/Volumes/Code_2TB/code/anthropic-lab' && claude resume claude-456",
   );
-  assert.equal(report.tabs[2]?.restartCommand, "cd '/Volumes/Code_2TB/code/collaboration'");
 });
 
 test("runWorkspaceCommands prints JSON with copyable commands", async () => {

package/packages/orp-workspace-launcher/test/tabs.test.js

@@ -111,16 +111,16 @@ test("buildWorkspaceTabsReport keeps duplicate titles unique and exposes generic
     "cd '/Volumes/Code_2TB/code/collaboration' && codex resume abc-123",
   );
   assert.equal(report.tabs[0]?.codexSessionId, "abc-123");
-  assert.equal(report.tabs[1]?.title, "anthropic-lab");
-  assert.equal(report.tabs[1]?.resumeCommand, "claude resume claude-456");
-  assert.equal(report.tabs[1]?.remoteBranch, "main");
+  assert.equal(report.tabs[1]?.title, "collaboration (2)");
+  assert.equal(report.tabs[1]?.codexSessionId, null);
+  assert.equal(report.tabs[2]?.title, "anthropic-lab");
+  assert.equal(report.tabs[2]?.resumeCommand, "claude resume claude-456");
+  assert.equal(report.tabs[2]?.remoteBranch, "main");
   assert.equal(
-    report.tabs[1]?.restartCommand,
+    report.tabs[2]?.restartCommand,
     "cd '/Volumes/Code_2TB/code/anthropic-lab' && claude resume claude-456",
   );
-  assert.equal(report.tabs[1]?.claudeSessionId, "claude-456");
-  assert.equal(report.tabs[2]?.title, "collaboration (2)");
-  assert.equal(report.tabs[2]?.codexSessionId, null);
+  assert.equal(report.tabs[2]?.claudeSessionId, "claude-456");
 });
 
 test("buildWorkspaceTabsReport ranks Codex tabs by recent local session activity", async () => {
@@ -183,6 +183,143 @@ test("buildWorkspaceTabsReport ranks Codex tabs by recent local session activity
   assert.equal(report.projects[2]?.path, "/Volumes/Code_2TB/code/no-session-project");
 });
 
+test("buildWorkspaceTabsReport ranks tracked Codex sessions by update time, not rollout creation time", async () => {
+  const tempDir = await makeTempDir();
+  const codexHome = path.join(tempDir, "codex-home");
+  const sessionsDir = path.join(codexHome, "sessions", "2026", "04", "15");
+  await fs.mkdir(sessionsDir, { recursive: true });
+
+  const oldRolloutUpdatedSessionId = "019d0000-0000-7000-8000-000000000021";
+  const newRolloutStaleSessionId = "019d0000-0000-7000-8000-000000000022";
+  const untrackedFreshSessionId = "019d0000-0000-7000-8000-000000000023";
+  const oldRolloutUpdatedPath = path.join(
+    sessionsDir,
+    `rollout-2026-04-15T01-00-00-${oldRolloutUpdatedSessionId}.jsonl`,
+  );
+  const newRolloutStalePath = path.join(
+    sessionsDir,
+    `rollout-2026-04-15T09-00-00-${newRolloutStaleSessionId}.jsonl`,
+  );
+  const untrackedFreshPath = path.join(sessionsDir, `rollout-2026-04-15T10-00-00-${untrackedFreshSessionId}.jsonl`);
+
+  await fs.writeFile(oldRolloutUpdatedPath, "{}\n", "utf8");
+  await fs.writeFile(newRolloutStalePath, "{}\n", "utf8");
+  await fs.writeFile(untrackedFreshPath, "{}\n", "utf8");
+  await fs.utimes(oldRolloutUpdatedPath, new Date("2026-04-15T11:00:00Z"), new Date("2026-04-15T11:00:00Z"));
+  await fs.utimes(newRolloutStalePath, new Date("2026-04-15T09:00:00Z"), new Date("2026-04-15T09:00:00Z"));
+  await fs.utimes(untrackedFreshPath, new Date("2026-04-15T12:00:00Z"), new Date("2026-04-15T12:00:00Z"));
+
+  const parsed = parseWorkspaceSource({
+    sourceType: "workspace-file",
+    sourceLabel: "/tmp/workspace.json",
+    title: "workspace",
+    workspaceManifest: {
+      version: "1",
+      workspaceId: "orp-main",
+      tabs: [
+        {
+          title: "new-rollout-stale",
+          path: "/Volumes/Code_2TB/code/new-rollout-stale",
+          resumeCommand: `codex resume ${newRolloutStaleSessionId}`,
+        },
+        {
+          title: "old-rollout-updated",
+          path: "/Volumes/Code_2TB/code/old-rollout-updated",
+          resumeCommand: `codex resume ${oldRolloutUpdatedSessionId}`,
+        },
+      ],
+    },
+    notes: "",
+  });
+
+  const report = buildWorkspaceTabsReport(
+    {
+      sourceType: "workspace-file",
+      sourceLabel: "/tmp/workspace.json",
+      title: "workspace",
+    },
+    parsed,
+    { codexHome },
+  );
+
+  assert.deepEqual(
+    report.tabs.map((tab) => tab.title),
+    ["old-rollout-updated", "new-rollout-stale"],
+  );
+});
+
+test("buildWorkspaceTabsReport bubbles a project when one attached Codex session is freshest", async () => {
+  const tempDir = await makeTempDir();
+  const codexHome = path.join(tempDir, "codex-home");
+  const sessionsDir = path.join(codexHome, "sessions", "2026", "04", "16");
+  await fs.mkdir(sessionsDir, { recursive: true });
+
+  const projectAOlderSessionId = "019d0000-0000-7000-8000-000000000011";
+  const projectANewerSessionId = "019d0000-0000-7000-8000-000000000012";
+  const projectBSessionId = "019d0000-0000-7000-8000-000000000013";
+
+  const projectAOlderPath = path.join(sessionsDir, `rollout-2026-04-16T01-00-00-${projectAOlderSessionId}.jsonl`);
+  const projectANewerPath = path.join(sessionsDir, `rollout-2026-04-16T03-00-00-${projectANewerSessionId}.jsonl`);
+  const projectBPath = path.join(sessionsDir, `rollout-2026-04-16T02-00-00-${projectBSessionId}.jsonl`);
+
+  await fs.writeFile(projectAOlderPath, "{}\n", "utf8");
+  await fs.writeFile(projectANewerPath, "{}\n", "utf8");
+  await fs.writeFile(projectBPath, "{}\n", "utf8");
+  await fs.utimes(projectAOlderPath, new Date("2026-04-16T01:00:00Z"), new Date("2026-04-16T01:00:00Z"));
+  await fs.utimes(projectANewerPath, new Date("2026-04-16T03:00:00Z"), new Date("2026-04-16T03:00:00Z"));
+  await fs.utimes(projectBPath, new Date("2026-04-16T02:00:00Z"), new Date("2026-04-16T02:00:00Z"));
+
+  const parsed = parseWorkspaceSource({
+    sourceType: "workspace-file",
+    sourceLabel: "/tmp/workspace.json",
+    title: "workspace",
+    workspaceManifest: {
+      version: "1",
+      workspaceId: "orp-main",
+      tabs: [
+        {
+          title: "project-b",
+          path: "/Volumes/Code_2TB/code/project-b",
+          resumeCommand: `codex resume ${projectBSessionId}`,
+        },
+        {
+          title: "project-a-old",
+          path: "/Volumes/Code_2TB/code/project-a",
+          resumeCommand: `codex resume ${projectAOlderSessionId}`,
+        },
+        {
+          title: "project-a-new",
+          path: "/Volumes/Code_2TB/code/project-a",
+          resumeCommand: `codex resume ${projectANewerSessionId}`,
+        },
+      ],
+    },
+    notes: "",
+  });
+
+  const report = buildWorkspaceTabsReport(
+    {
+      sourceType: "workspace-file",
+      sourceLabel: "/tmp/workspace.json",
+      title: "workspace",
+    },
+    parsed,
+    { codexHome },
+  );
+
+  assert.deepEqual(
+    report.tabs.map((tab) => tab.title),
+    ["project-a-new", "project-a-old", "project-b"],
+  );
+  assert.equal(report.projects[0]?.path, "/Volumes/Code_2TB/code/project-a");
+  assert.equal(report.projects[0]?.sessionCount, 2);
+  assert.deepEqual(
+    report.projects[0]?.sessions.map((session) => session.title),
+    ["project-a-new", "project-a-old"],
+  );
+  assert.equal(report.projects[1]?.path, "/Volumes/Code_2TB/code/project-b");
+});
+
 test("runWorkspaceTabs prints JSON without launch commands", async () => {
   const tempDir = await makeTempDir();
   const manifestPath = path.join(tempDir, "workspace.json");