How long does the integration take?

Change one base URL in an OpenAI-compatible client, or install the Python or TypeScript SDK; the quickstart routes a first request in a few minutes.

Will this lock us into another vendor?

No. The router is Apache-2.0 open source, your provider keys stay yours, and the API is OpenAI-compatible, so leaving is a change of base URL.

What happens when a provider fails?

The router falls back to the next model your policy allows and records the fallback in the trace, so an outage costs a retry, not an incident.

Pre-release 3.0.1 M2 Ghidra Follow-up - Prompt

Name: Pre-release 3.0.1 M2 Ghidra Follow-up
Author: akspa0

Prompt file imported from akspa0/parp-tools (.codex/prompts/pre-release-3-0-1-m2-ghidra-followup.md). Copyright stays with the author.

Continue the wow.exe reverse-engineering pass for pre-release 3.0.1 M2 handling.

Read First

gillijimproject_refactor/documentation/pre-release-3.0.1-m2-wow-exe-guide.md
.codex/prompts/pre-release-m2-rendering-recovery.md
gillijimproject_refactor/src/MdxViewer/memory-bank/activeContext.md

Goal

Resolve the remaining unknowns in the pre-release 3.0.1 model path directly from the binary.

This prompt is for when implementation is blocked by missing structure or loader facts, not for speculative code edits.

Required Investigation Areas

Reconfirm the shared loader chain:
- FUN_0077e2c0
- FUN_0077d3c0
- FUN_0079bc70
- FUN_0079bc50
- FUN_0079bb30
- FUN_0079a8c0
Map unresolved record families one at a time:
- 0x70
- 0x2C
- 0x38
- 0xD4
- 0x7C
- legacy 0xDC and 0x1F8
- later 0xE0 and 0x234
If useful, rename key functions or add disassembly comments so future chats do not have to rediscover them.
Keep Track B separate unless the binary clearly proves shared renderer behavior is also involved.

Required Principles

Prefer call-graph and xref evidence over guessed struct names.
Record exact record sizes and branch points before naming fields.
If a function meaning is still unclear, say that clearly instead of inventing a label.
Update the guide when new facts become high confidence.

Deliverables

Return all items:

New high-confidence findings
Exact function addresses and inferred roles
Any renames or comments applied in Ghidra
Which unresolved family is now understood and which still is not
Concrete implementation implications for MdxViewer
Guide or memory-bank updates required

First Output

Start with:

the unresolved part you will map first
the exact function addresses you will inspect
whether the goal is implementation support, loader tracing, or field naming

Install Pre-release 3.0.1 M2 Ghidra Follow-up and it becomes one more candidate the router can pick - when it fits.

A listing is a routing target with a manifest: what it does, which domains it covers, what it costs and who publishes it. Once installed it sits beside your own models and tools, is scored like any other candidate for each request, and shows up in the trace when it wins. Ratings come from workspaces that installed it, one per account.

Continue the `wow.exe` reverse-engineering pass for pre-release `3.0.1` M2 handling. ## Read First 1. `gillijimproject_refactor/documentation/pre-release-3.0.1-m2-wow-exe-guide.md` 2. `.codex/prompts/pre-release-m2-rendering-recovery.md` 3. `gillijimproject_refactor/src/MdxViewer/memory-bank/activeContext.md` ## Goal Resolve the remaining unknowns in the pre-release `3.0.1` model path directly from the binary. This prompt is for when implementation is blocked by missing structure or loader facts, not for speculative code edits. ## Required Investigation Areas 1. Reconfirm the shared loader chain: - `FUN_0077e2c0` - `FUN_0077d3c0` - `FUN_0079bc70` - `FUN_0079bc50` - `FUN_0079bb30` - `FUN_0079a8c0` 2. Map unresolved record families one at a time: - `0x70` - `0x2C` - `0x38` - `0xD4` - `0x7C` - legacy `0xDC` and `0x1F8` - later `0xE0` and `0x234` 3. If useful, rename key functions or add disassembly comments so future chats do not have to rediscover them. 4. Keep Track B separate unless the binary clearly proves shared renderer behavior is also involved. ## Required Principles - Prefer call-graph and xref evidence over guessed struct names. - Record exact record sizes and branch points before naming fields. - If a function meaning is still unclear, say that clearly instead of inventing a label. - Update the guide when new facts become high confidence. ## Deliverables Return all items: 1. New high-confidence findings 2. Exact function addresses and inferred roles 3. Any renames or comments applied in Ghidra 4. Which unresolved family is now understood and which still is not 5. Concrete implementation implications for `MdxViewer` 6. Guide or memory-bank updates required ## First Output Start with: 1. the unresolved part you will map first 2. the exact function addresses you will inspect 3. whether the goal is implementation support, loader tracing, or field naming

{ "prompt": "Continue the `wow.exe` reverse-engineering pass for pre-release `3.0.1` M2 handling.\n\n## Read First\n\n1. `gillijimproject_refactor/documentation/pre-release-3.0.1-m2-wow-exe-guide.md`\n2. `.codex/prompts/pre-release-m2-rendering-recovery.md`\n3. `gillijimproject_refactor/src/MdxViewer/memory-bank/activeContext.md`\n\n## Goal\n\nResolve the remaining unknowns in the pre-release `3.0.1` model path directly from the binary.\n\nThis prompt is for when implementation is blocked by missing structure or loader facts, not for speculative code edits.\n\n## Required Investigation Areas\n\n1. Reconfirm the shared loader chain:\n - `FUN_0077e2c0`\n - `FUN_0077d3c0`\n - `FUN_0079bc70`\n - `FUN_0079bc50`\n - `FUN_0079bb30`\n - `FUN_0079a8c0`\n2. Map unresolved record families one at a time:\n - `0x70`\n - `0x2C`\n - `0x38`\n - `0xD4`\n - `0x7C`\n - legacy `0xDC` and `0x1F8`\n - later `0xE0` and `0x234`\n3. If useful, rename key functions or add disassembly comments so future chats do not have to rediscover them.\n4. Keep Track B separate unless the binary clearly proves shared renderer behavior is also involved.\n\n## Required Principles\n\n- Prefer call-graph and xref evidence over guessed struct names.\n- Record exact record sizes and branch points before naming fields.\n- If a function meaning is still unclear, say that clearly instead of inventing a label.\n- Update the guide when new facts become high confidence.\n\n## Deliverables\n\nReturn all items:\n\n1. New high-confidence findings\n2. Exact function addresses and inferred roles\n3. Any renames or comments applied in Ghidra\n4. Which unresolved family is now understood and which still is not\n5. Concrete implementation implications for `MdxViewer`\n6. Guide or memory-bank updates required\n\n## First Output\n\nStart with:\n\n1. the unresolved part you will map first\n2. the exact function addresses you will inspect\n3. whether the goal is implementation support, loader tracing, or field naming", "variables": [], "notes": "Continue Ghidra-backed reverse engineering for pre-release 3.0.1 M2 files when field mapping, validator order, or loader behavior is still unresolved. Arguments: Optional function family, failing asset, or unresolved structure size to focus on.", "metadata": { "source": { "provider": "github-codex-prompts", "repository": "https://github.com/akspa0/parp-tools", "path": ".codex/prompts/pre-release-3-0-1-m2-ghidra-followup.md", "ref": "4f7fb94cfad6bf6348c55c28ae9361c2d2fd9cb7", "url": "https://github.com/akspa0/parp-tools/blob/4f7fb94cfad6bf6348c55c28ae9361c2d2fd9cb7/.codex/prompts/pre-release-3-0-1-m2-ghidra-followup.md", "key": "akspa0/parp-tools/.codex/prompts/pre-release-3-0-1-m2-ghidra-followup.md" } } }

Pre-release 3.0.1 M2 Ghidra Follow-up

About

Read First

Goal

Required Investigation Areas

Required Principles

Deliverables

First Output

Install Pre-release 3.0.1 M2 Ghidra Follow-up and it becomes one more candidate the router can pick - when it fits.

What it costs to adopt - in engineering time.

One base URL

No vendor lock-in

Failover built in

Use it

Manifest

Reviews