docs/.gitignore (1)

17-19: Optional ignores for local developer artefacts

Consider ignoring common local files to reduce accidental commits.

 # OS files
 .DS_Store
 Thumbs.db
+.python-version
+.env
+.env.*

docs/content/development/setup.md (1)

343-348: Prefer graceful termination before SIGKILL

Suggest using kill -TERM <PID> and only falling back to kill -9 if needed, to allow clean-up hooks (useful during dev when files or ports need graceful release).

-# Kill process
-kill -9 <PID>
+# Terminate gracefully
+kill -TERM <PID>
+# If the process doesn’t exit within a few seconds, force kill:
+kill -9 <PID>

docs/content/api-reference/ollama.md (2)

76-83: Add a language to fenced code block and clarify content type

This streaming example is line-delimited JSON (NDJSON). Specify a fence language and note the response Content-Type to help readers consume it correctly.

-```
+```json
 {"model":"llama3.2:latest","created_at":"2024-01-15T10:30:00Z","response":"The","done":false}
 {"model":"llama3.2:latest","created_at":"2024-01-15T10:30:00Z","response":" sky","done":false}
 {"model":"llama3.2:latest","created_at":"2024-01-15T10:30:00Z","response":" appears","done":false}
 ...
 {"model":"llama3.2:latest","created_at":"2024-01-15T10:30:01Z","response":"","done":true,"total_duration":5250000000}

+Note: Streaming responses are delivered as newline-delimited JSON (Content-Type: application/x-ndjson).

---

`191-200`: **Add a language to fenced code block (markdownlint MD040)**

The SSE-style stream should declare a language for readability and linting.

```diff
-```
+```text
 data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","created":1705334400,"model":"llama3.2:latest","choices":[{"index":0,"delta":{"role":"assistant","content":"2"},"finish_reason":null}]}
 
 data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","created":1705334400,"model":"llama3.2:latest","choices":[{"index":0,"delta":{"content":" +"},"finish_reason":null}]}
 
 data: {"id":"chatcmpl-abc123","object":"chat.completion.chunk","created":1705334400,"model":"llama3.2:latest","choices":[{"index":0,"delta":{"content":" 2"},"finish_reason":null}]}
 
 data: [DONE]

+Note: Streaming responses use Server-Sent Events (Content-Type: text/event-stream).

</blockquote></details>
<details>
<summary>docs/content/api-reference/openai.md (2)</summary><blockquote>

`121-131`: **Add a language to fenced code block (markdownlint MD040)**

Declare a language for the SSE stream example.

```diff
-```
+```text
 data: {"id":"chatcmpl-8q9ABC123","object":"chat.completion.chunk","created":1705334400,"model":"gpt-3.5-turbo-0125","system_fingerprint":"fp_abc123","choices":[{"index":0,"delta":{"role":"assistant","content":"Quantum"},"logprobs":null,"finish_reason":null}]}
 
 data: {"id":"chatcmpl-8q9ABC123","object":"chat.completion.chunk","created":1705334400,"model":"gpt-3.5-turbo-0125","system_fingerprint":"fp_abc123","choices":[{"index":0,"delta":{"content":" computing"},"logprobs":null,"finish_reason":null}]}
 
 ...
 
 data: {"id":"chatcmpl-8q9ABC123","object":"chat.completion.chunk","created":1705334400,"model":"gpt-3.5-turbo-0125","system_fingerprint":"fp_abc123","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}]}
 
 data: [DONE]

---

329-334: Consistent Australian English: behaviour

Per repository guidelines, prefer Australian English in docs.

-| `function_call` | string/object | Function calling behavior |
+| `function_call` | string/object | Function calling behaviour |
@@
-| `tool_choice` | string/object | Tool selection behavior |
+| `tool_choice` | string/object | Tool selection behaviour |

docs/content/configuration/practices/performance.md (1)

2-5: Align wording with Australian English across front matter

Per guidelines, use Australian English consistently.

-title: "Olla Performance Best Practices - Optimization & Tuning Guide"
-description: "Complete performance optimization guide for Olla. Engine selection, connection pooling, memory tuning, streaming configuration and production optimizations."
-keywords: ["olla performance", "performance optimization", "performance tuning", "connection pooling", "memory optimization", "streaming performance"]
+title: "Olla Performance Best Practices - Optimisation & Tuning Guide"
+description: "Complete performance optimisation guide for Olla. Engine selection, connection pooling, memory tuning, streaming configuration and production optimisations."
+keywords: ["olla performance", "performance optimisation", "performance tuning", "connection pooling", "memory optimisation", "streaming performance"]

docs/content/development/contributing.md (1)

66-72: Add language to fenced code block (markdownlint MD040).

Specify a language for this fence to satisfy linting and improve rendering.

-```
+```text
 type(scope): description

 [optional body]

 [optional footer]

</blockquote></details>
<details>
<summary>docs/content/development/patterns.md (1)</summary><blockquote>

`165-173`: **Clarify backpressure behaviour in Submit.**

Default case silently drops/blocks undefined. Either block, drop with metrics, or expose a non-blocking SubmitTry API.


Example:

- Block: remove default to naturally block on full channel.
- Drop with metric/log: increment a counter and return.
- Provide `TrySubmit` returning bool.

</blockquote></details>
<details>
<summary>docs/content/development/benchmarking.md (1)</summary><blockquote>

`316-342`: **Consider using a stable Go version or matrix in CI sample.**

Hard-coding an unreleased version can confuse readers. Suggest ‘stable’ or a tested matrix.


```diff
-      - uses: actions/setup-go@v4
-        with:
-          go-version: '1.24'
+      - uses: actions/setup-go@v4
+        with:
+          go-version: 'stable'

docs/content/faq.md (3)

58-66: Heading punctuation: add a comma for readability.

Improves clarity and matches guidance.

-### What is stream_buffer_size and how should I tune it?
+### What is stream_buffer_size, and how should I tune it?

---

314-318: Hyphenate compound modifier (“auto-proxy profile”).

Australian English style; improves precision.

-### Should I use auto proxy profile?
+### Should I use the auto-proxy profile?

---

322-325: Use en dashes for numeric ranges.

Typographic polish for ranges.

-- **Production**: 30-60 seconds
-- **Development**: 10-30 seconds
-- **Critical systems**: 5-10 seconds
+- **Production**: 30–60 seconds
+- **Development**: 10–30 seconds
+- **Critical systems**: 5–10 seconds

docs/content/development/testing.md (1)

20-30: Add language to fenced code block (markdownlint MD040).

Label the tree listing for consistency and lint compliance.

-```
+```text
 .
 ├── internal/
 │   └── */                  # Unit tests alongside code
 │       └── *_test.go
 ├── test/
 │   ├── integration/        # Integration tests
 │   ├── scripts/           # Test scripts
 │   └── fixtures/          # Test data
 └── benchmarks/            # Performance benchmarks

</blockquote></details>
<details>
<summary>docs/content/configuration/reference.md (1)</summary><blockquote>

`353-359`: **Minor style: introduce log levels with a sentence.**

Prevents a “loose punctuation” feel and reads better.


```diff
-Log levels:
+The following log levels are supported:
 
 - `debug`: Detailed debugging information
 - `info`: Normal operational messages
 - `warn`: Warning conditions
 - `error`: Error conditions only

docs/content/getting-started/installation.md (6)

95-96: Fix broken/incorrect “Architecture Overview” link

The target likely lives under Concepts, not Development. Please verify the path and update to keep the IA consistent.

Proposed change:

- [Architecture Overview](../development/architecture.md) - Learn how Olla works
+ [Architecture Overview](../concepts/overview.md) - Learn how Olla works

---

113-113: Validate Troubleshooting link

The link to contributing/troubleshooting may be incorrect or outdated. Please confirm that the document exists at this path; otherwise update to the correct target.

- For more help, check the [troubleshooting guide](../development/contributing.md#troubleshooting) or [open an issue](https://github.com/thushan/olla/issues).
+ For more help, check the [troubleshooting guide](../contributing/troubleshooting.md) or [open an issue](https://github.com/thushan/olla/issues).

---

39-41: Capitalise proper nouns in Docker section

Use consistent capitalisation for product names.

- # for locally installed lmstudio, ollama or vllm
+ # for locally installed LM Studio, Ollama, or vLLM

---

104-106: Minor wording: executable permission

“Execute permissions” reads a bit off. “Executable permission(s)” is clearer.

- : On Linux/macOS, ensure the binary has execute permissions: `chmod +x olla`
+ : On Linux/macOS, ensure the binary has executable permission: `chmod +x olla`

---

80-89: markdownlint MD046: code block style

The linter expects indented code blocks here, but the doc consistently uses fenced blocks elsewhere. Prefer consistency: either switch this block to indented style or relax MD046 in your markdownlint config to allow fenced blocks.

If keeping fenced blocks (recommended), update markdownlint config:

• Disable MD046 or set style to “fenced”.

---

20-31: Windows install path is missing

Requirements list mentions Windows, but the primary install script and examples target Linux/macOS. Consider adding a Windows subsection (PowerShell) so users don’t guess.

Example:

# Windows PowerShell
iwr -useb https://raw.githubusercontent.com/thushan/olla/main/install.ps1 | iex

docs/content/api-reference/lmstudio.md (1)

125-137: Add language to fenced block (markdownlint MD040)

Specify a language for the streaming SSE snippet to satisfy the linter and improve rendering.

-```
+```text
 data: {"id":"chatcmpl-lmstudio-abc123","object":"chat.completion.chunk","created":1705334400,"model":"phi-3-mini","choices":[{"index":0,"delta":{"role":"assistant","content":"Here's"},"finish_reason":null}]}
 ...
 data: [DONE]

</blockquote></details>
<details>
<summary>docs/content/integrations/frontend/openwebui.md (3)</summary><blockquote>

`11-11`: **Typo: GitHub capitalization**

```diff
-You can find an example integration ... see [latest in Github](
+You can find an example integration ... see [latest in GitHub](

---

355-357: Ping may be unavailable in containers

open-webui images often don’t include ping. A curl to Olla’s health endpoint is more reliable.

-   docker exec openwebui ping olla
+   docker exec openwebui curl -sf http://olla:40114/internal/health

---

60-76: Add language to diagram code fence

For consistency and to satisfy linters, mark the ASCII diagram as text.

-```
+```text
 # diagram …

</blockquote></details>
<details>
<summary>docs/content/development/circuit-breaker.md (2)</summary><blockquote>

`42-59`: **Replace bolded labels with proper headings (markdownlint MD036)**

Use headings for the three states to improve structure and navigation.

```diff
-**Closed (Normal Operation)**
+#### Closed (Normal operation)

-**Open (Circuit Tripped)**
+#### Open (Circuit tripped)

-**Half-Open (Testing Recovery)**
+#### Half-open (testing recovery)

---

131-133: Minor grammar

Add a comma for readability.

-The `/internal/status` endpoint shows endpoint health which reflects circuit breaker state:
+The `/internal/status` endpoint shows endpoint health, which reflects circuit breaker state:

docs/content/api-reference/vllm.md (3)

153-163: Add language to streaming SSE block (markdownlint MD040)

-```
+```text
 data: {"id":"chatcmpl-vllm-abc123","object":"chat.completion.chunk", ...}
 ...
 data: [DONE]

---

`266-286`: **Add language to Prometheus metrics block (markdownlint MD040)**

```diff
-```
+```text
 # HELP vllm_request_duration_seconds Request duration in seconds
 # TYPE vllm_request_duration_seconds histogram
 ...

---

`329-337`: **Australian English spelling**

Prefer “optimisations” over “optimizations”.

```diff
-vLLM provides several performance optimizations:
+vLLM provides several performance optimisations:

docs/content/concepts/health-checking.md (4)

39-46: Add language to diagram code fence (markdownlint MD040)

-```
+```text
 ┌─────────────┐     ┌──────────────┐     ┌─────────────┐
 ...

---

`178-186`: **Add language to circuit diagram code fence (markdownlint MD040)**

```diff
-```
+```text
      Closed (Normal)
 ...

---

`107-111`: **Use en dashes for numeric ranges (en‑AU style)**

```diff
-**Local endpoints**: 2-5 seconds
-**LAN endpoints**: 5-10 seconds  
-**Remote/Cloud**: 15-30 seconds
-**Critical endpoints**: 2-3 seconds
+**Local endpoints**: 2–5 seconds
+**LAN endpoints**: 5–10 seconds
+**Remote/Cloud**: 15–30 seconds
+**Critical endpoints**: 2–3 seconds

---

172-200: Align circuit-breaker values across docs

This page states “3 test requests” and “2 successes to close.” The dedicated circuit-breaker page currently mixes 1 vs 3 requests. Please standardise the thresholds across both pages (and ideally derive from a single source of truth or include snippet) to avoid operator confusion.

docs/content/stylesheets/extra.css (1)

29-31: Accessibility: ensure sufficient contrast

White text on the accent-coloured header and controls may fail WCAG contrast in some displays. Consider verifying contrast ratios and, if needed, darken the accent for header areas or reduce white usage in buttons/links.

Also applies to: 95-105

docs/content/integrations/overview.md (1)

9-9: Minor grammar

“front‑ends integrations” reads awkwardly. Suggest:

-Olla supports various backends (endpoints) and front-ends integrations powered by Olla [Profiles](../concepts/profile-system.md).
+Olla supports various backends (endpoints) and front‑end integrations powered by Olla [Profiles](../concepts/profile-system.md).

docs/mkdocs.yml (2)

161-161: Add newline at end of file.

Some tools (and yamllint) expect a trailing newline.

No content change needed; just ensure the file ends with a newline.

---

11-60: Consider enabling strict mode to catch broken links during CI.

MkDocs strict mode fails the build on bad links/refs; useful for larger docs sets being published.

Apply this at top-level (e.g., after docs_dir):

 docs_dir: content
+strict: true

docs/content/development/overview.md (4)

13-13: Comma before “but” in compound sentence.

Prefer: “…on Linux and macOS, you can develop on Windows, but you may hit…”.

-Primary development has been done on Linux and macOS, you can develop on Windows but you may hit UAC prompts for port usage and occasional pauses at startup.
+Primary development has been done on Linux and macOS; you can develop on Windows, but you may hit UAC prompts for port usage and occasional pauses at startup.

---

35-41: Add a language to fenced code block (markdownlint MD040).

Use “text” for ASCII trees to satisfy linters.

-```
+```text
 internal/
 ├── core/           # Domain layer - business logic, zero external dependencies
 ├── adapter/        # Infrastructure - implementations of core interfaces
 └── app/            # Application layer - HTTP handlers, orchestration

---

`91-103`: **Add language to fenced code block (markdownlint MD040).**

Use “text” for directory layouts.

```diff
-```
+```text
 .
 ├── cmd/            # Application entry points
 ├── internal/       # Private application code
 │   ├── core/      # Business logic
 │   ├── adapter/   # External integrations
 │   └── app/       # Application layer
 ├── pkg/           # Public packages
 ├── config/        # Configuration files
 ├── test/          # Integration tests
 └── docs/          # Documentation

---

`121-124`: **Use determiners with superlatives.**

Prefer “the highest priority endpoint” and “the least busy endpoint”.

```diff
-- **Priority**: Routes to highest priority endpoint
+- **Priority**: Routes to the highest priority endpoint
-- **Least-Connections**: Routes to least busy endpoint
+- **Least-Connections**: Routes to the least busy endpoint

Also applies to: 123-124

docs/content/configuration/overview.md (3)

25-26: Minor punctuation for readability.

Add a comma after “behaviour” clause.

-Olla uses a YAML configuration file to control all aspects of its behaviour. This page provides an overview of the configuration structure with links to detailed documentation for each section.
+Olla uses a YAML configuration file to control all aspects of its behaviour. This page provides an overview of the configuration structure, with links to detailed documentation for each section.

---

166-173: Use determiners with superlatives.

Use “the highest priority endpoint” and “the least busy endpoint”.

-| **priority** | Routes to highest priority endpoint | Preferring local/cheaper endpoints |
+| **priority** | Routes to the highest priority endpoint | Preferring local/cheaper endpoints |
-| **least-connections** | Routes to least busy endpoint | Optimising response times |
+| **least-connections** | Routes to the least busy endpoint | Optimising response times |

Also applies to: 170-172

---

103-109: Compound adjectives hyphenation (optional).

“Health-aware routing” and “client-IP detection” can be hyphenated when used adjectivally. Current wording is still clear; consider only if you prefer stricter style.

No code changes required.

docs/content/concepts/overview.md (1)

24-27: Use determiners with superlatives.

Prefer “the least busy endpoints”.

-- **Least-connections**: Routes to least busy endpoints
+- **Least-connections**: Routes to the least busy endpoints

docs/content/getting-started/quickstart.md (1)

69-76: Minor grammar tweak.

Add “the” for smoother phrasing.

-List available models through the proxy:
+List available models through the proxy:

Note: The wording is already acceptable; keep if you prefer the current phrasing.

docs/content/integrations/backend/vllm.md (4)

3-5: Use Australian English consistently in prose.

Adopt “optimisation” (not “optimization”) in descriptions. Keep CLI flags as-is.

-description: Configure vLLM with Olla proxy for production-grade LLM serving. PagedAttention, continuous batching, tensor parallelism, and GPU optimization.
+description: Configure vLLM with Olla proxy for production-grade LLM serving. PagedAttention, continuous batching, tensor parallelism, and GPU optimisation.

---

280-286: Standardise “-isation” spellings in prose.

Use en‑AU variants in explanatory text: “tokenisation/quantisation/optimisation”. Keep endpoint paths and CLI flags (tokenize/detokenize/--quantization) unchanged.

Examples:

• “Quantisation Support” (already correct)
• “Tokenisation API” (already correct)
• Ensure surrounding sentences use “quantisation” even when a flag shows “--quantization”.

Also applies to: 287-301

---

433-441: Use neutral phrasing for TTFT note and keep en‑AU.

Prefer “long time to first token (TTFT)” over “High time…”, and keep “utilisation”.

-**Issue**: High time to first token (TTFT)
+**Issue**: Long time to first token (TTFT)
@@
-2. Increase GPU memory utilisation
+2. Increase GPU memory utilisation
-3. Use smaller model or quantisation
+3. Use a smaller model or quantisation

---

427-432: Consistent en‑AU spelling while preserving CLI flag names.

Keep prose as “quantisation”, keep the flag as “--quantization”.

-3. Use quantisation (`--quantization awq` or `--quantization gptq`)
+3. Use quantisation (`--quantization awq` or `--quantization gptq`)

(Current line already correct for en‑AU; ensure there’s no mixed “quantization” in prose elsewhere.)

docs/content/development/architecture.md (2)

599-599: Add hyphen to compound adjective.

The compound adjective "worst case" should be hyphenated when modifying a noun:

-1. **Endpoint Selection**: O(1) for priority, O(n) worst case
+1. **Endpoint Selection**: O(1) for priority, O(n) worst-case

---

62-90: Add language specification to code block.

The ASCII diagram code block should specify a language for proper rendering:

-```
+```text
 ┌─────────────────────────────────────────────────────────┐

docs/content/api-reference/overview.md (1)

13-15: Add language specification to code block.

The base URL code block should specify a language for proper rendering:

-```
+```text
 http://localhost:40114

</blockquote></details>
<details>
<summary>docs/content/concepts/profile-system.md (2)</summary><blockquote>

`610-610`: **Add comma before interrupter.**

The word "however" as an interrupter should be preceded by a comma:

```diff
-Don't update versions in natively supported profiles however.
+Don't update versions in natively supported profiles, however.

---

108-123: Convert indented code block to fenced format.

The code block should use fenced format instead of indentation:

-    ```
-    ROUTE                             | METHOD | DESCRIPTION
-    /olla/lmstudio/v1/models          | GET    | lmstudio models (OpenAI format)
-    ...
-    ```
+```text
+ROUTE                             | METHOD | DESCRIPTION
+/olla/lmstudio/v1/models          | GET    | lmstudio models (OpenAI format)
+...
+```

docs/content/index.md (2)

124-129: Convert indented code blocks to fenced format.

The code blocks should use fenced format instead of indentation:

-=== "Using Go"
-
-    ```bash
-    go install github.com/thushan/olla@latest
-    olla
-    ```
+=== "Using Go"
+
+```bash
+go install github.com/thushan/olla@latest
+olla
+```

---

131-138: Convert indented code blocks to fenced format.

The code blocks should use fenced format instead of indentation:

-=== "From Source"
-
-    ```bash
-    git clone https://github.com/thushan/olla.git
-    cd olla
-    make build-release
-    ./olla
-    ```
+=== "From Source"
+
+```bash
+git clone https://github.com/thushan/olla.git
+cd olla
+make build-release
+./olla
+```

docs/content/concepts/model-unification.md (1)

398-398: Add comma in compound sentence.

The compound sentence needs a comma before "but":

-Models with same name but different actual files may be incorrectly unified
+Models with same name, but different actual files may be incorrectly unified

docs/content/concepts/load-balancing.md (3)

10-12: Grammar: add missing determiners and hyphenation for clarity

Use “the” with superlatives and hyphenate compound adjectives.

- - `priority` _(default)_ - Routes to highest priority endpoint
+ - `priority` _(default)_ - Routes to the highest-priority endpoint
  - `round-robin` - Cycles through endpoints evenly
- - `least-connections` - Routes to least busy endpoint
+ - `least-connections` - Routes to the least-busy endpoint

---

35-39: Grammar: add missing determiner and hyphenation

Superlative needs “the”; hyphenate for readability.

-2. Selects from highest priority tier only
+2. Selects from the highest-priority tier only

---

110-114: Grammar: add missing determiner

Add “the” before “fewest”.

-Routes to the endpoint with fewest active connections.
+Routes to the endpoint with the fewest active connections.

docs/content/configuration/practices/monitoring.md (2)

245-249: Clarify latency guidance vs availability of percentiles

You set thresholds for P50/P95/P99 but note that percentiles aren’t exposed by the status endpoint. Explicitly state that these should be computed in the monitoring system (e.g. Prometheus histogram) rather than fetched from Olla.

Also applies to: 103-110

---

437-443: Prefer “why” comments for Go snippet

Shift to a “why” comment to match style guidance and help readers understand the intent.

-// Wrap handlers with OpenTelemetry
+// Add tracing to attribute proxy time vs backend latency, enabling root-cause analysis
 import "go.opentelemetry.io/otel"
 
 tracer := otel.Tracer("olla")
 ctx, span := tracer.Start(ctx, "proxy_request")
 defer span.End()

docs/content/integrations/backend/ollama.md (4)

294-299: Fix markdown lint: specify language for fenced block

Add a language identifier to satisfy MD040 and improve rendering.

-```
+```text
 model:tag
 model:version
 namespace/model:tag

---

`323-339`: **Mark commented JSON as jsonc to avoid invalid JSON copy-paste errors**

The example includes comments; jsonc is more accurate and prevents confusion.


```diff
-```json
+```jsonc
 {
   "options": {
     "temperature": 0.8,      // Randomness (0-1)

---

649-658: Align endpoint path with the rest of the docs

Elsewhere the docs use /internal/stats/models. Update here for consistency (unless the service exposes both).

-# Monitor model availability
-curl http://localhost:40114/internal/status/models
+# Monitor model availability
+curl http://localhost:40114/internal/stats/models

---

378-405: Compose vs Swarm: GPU settings under deploy won’t apply in plain docker-compose

docker-compose ignores the deploy section (it’s Swarm-only). Consider adding a note or an alternative example using device_requests for Compose.

Example alternative (Compose):

services:
  ollama:
    image: ollama/ollama
    deploy: {}  # ignored by docker-compose
    device_cgroup_rules:
      - 'c 195:* rmw'   # or use device_requests with compose v2.10+
    deploy:
      resources: {}     # only active in Swarm

Or:

services:
  ollama:
    image: ollama/ollama
    deploy: {}
    gpu: all  # Docker Desktop convenience flag

Add a brief note explaining why this matters to prevent confusion.

docs/content/integrations/backend/lmstudio.md (1)

372-382: Optional: update LangChain example to current packages

LangChain’s OpenAI bindings have moved; many users now rely on langchain_openai. Consider updating or adding a second snippet to reduce friction for new setups.

Proposed alternative:

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="http://localhost:40114/olla/lm-studio/v1",
    api_key="not-needed",
    model="mistral-7b-instruct"
)

If your audience targets older versions, keep both with a note.

Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro

Learnt from: CR
PR: thushan/olla#0
File: CLAUDE.md:0-0
Timestamp: 2025-08-11T11:48:06.612Z
Learning: Applies to internal/adapter/proxy/**/*_test.go : Maintain shared proxy tests to ensure both Sherpa and Olla engines remain compatible